Add files using upload-large-folder tool

Orient_Anything/.gitignore

@@ -0,0 +1,4 @@
+**/__pycache__
+.gradio
+.locks
+models*

Orient_Anything/LICENSE

@@ -0,0 +1,395 @@
+Attribution 4.0 International
+
+=======================================================================
+
+Creative Commons Corporation ("Creative Commons") is not a law firm and
+does not provide legal services or legal advice. Distribution of
+Creative Commons public licenses does not create a lawyer-client or
+other relationship. Creative Commons makes its licenses and related
+information available on an "as-is" basis. Creative Commons gives no
+warranties regarding its licenses, any material licensed under their
+terms and conditions, or any related information. Creative Commons
+disclaims all liability for damages resulting from their use to the
+fullest extent possible.
+
+Using Creative Commons Public Licenses
+
+Creative Commons public licenses provide a standard set of terms and
+conditions that creators and other rights holders may use to share
+original works of authorship and other material subject to copyright
+and certain other rights specified in the public license below. The
+following considerations are for informational purposes only, are not
+exhaustive, and do not form part of our licenses.
+
+     Considerations for licensors: Our public licenses are
+     intended for use by those authorized to give the public
+     permission to use material in ways otherwise restricted by
+     copyright and certain other rights. Our licenses are
+     irrevocable. Licensors should read and understand the terms
+     and conditions of the license they choose before applying it.
+     Licensors should also secure all rights necessary before
+     applying our licenses so that the public can reuse the
+     material as expected. Licensors should clearly mark any
+     material not subject to the license. This includes other CC-
+     licensed material, or material used under an exception or
+     limitation to copyright. More considerations for licensors:
+    wiki.creativecommons.org/Considerations_for_licensors
+
+     Considerations for the public: By using one of our public
+     licenses, a licensor grants the public permission to use the
+     licensed material under specified terms and conditions. If
+     the licensor's permission is not necessary for any reason--for
+     example, because of any applicable exception or limitation to
+     copyright--then that use is not regulated by the license. Our
+     licenses grant only permissions under copyright and certain
+     other rights that a licensor has authority to grant. Use of
+     the licensed material may still be restricted for other
+     reasons, including because others have copyright or other
+     rights in the material. A licensor may make special requests,
+     such as asking that all changes be marked or described.
+     Although not required by our licenses, you are encouraged to
+     respect those requests where reasonable. More considerations
+     for the public:
+    wiki.creativecommons.org/Considerations_for_licensees
+
+=======================================================================
+
+Creative Commons Attribution 4.0 International Public License
+
+By exercising the Licensed Rights (defined below), You accept and agree
+to be bound by the terms and conditions of this Creative Commons
+Attribution 4.0 International Public License ("Public License"). To the
+extent this Public License may be interpreted as a contract, You are
+granted the Licensed Rights in consideration of Your acceptance of
+these terms and conditions, and the Licensor grants You such rights in
+consideration of benefits the Licensor receives from making the
+Licensed Material available under these terms and conditions.
+
+
+Section 1 -- Definitions.
+
+  a. Adapted Material means material subject to Copyright and Similar
+     Rights that is derived from or based upon the Licensed Material
+     and in which the Licensed Material is translated, altered,
+     arranged, transformed, or otherwise modified in a manner requiring
+     permission under the Copyright and Similar Rights held by the
+     Licensor. For purposes of this Public License, where the Licensed
+     Material is a musical work, performance, or sound recording,
+     Adapted Material is always produced where the Licensed Material is
+     synched in timed relation with a moving image.
+
+  b. Adapter's License means the license You apply to Your Copyright
+     and Similar Rights in Your contributions to Adapted Material in
+     accordance with the terms and conditions of this Public License.
+
+  c. Copyright and Similar Rights means copyright and/or similar rights
+     closely related to copyright including, without limitation,
+     performance, broadcast, sound recording, and Sui Generis Database
+     Rights, without regard to how the rights are labeled or
+     categorized. For purposes of this Public License, the rights
+     specified in Section 2(b)(1)-(2) are not Copyright and Similar
+     Rights.
+
+  d. Effective Technological Measures means those measures that, in the
+     absence of proper authority, may not be circumvented under laws
+     fulfilling obligations under Article 11 of the WIPO Copyright
+     Treaty adopted on December 20, 1996, and/or similar international
+     agreements.
+
+  e. Exceptions and Limitations means fair use, fair dealing, and/or
+     any other exception or limitation to Copyright and Similar Rights
+     that applies to Your use of the Licensed Material.
+
+  f. Licensed Material means the artistic or literary work, database,
+     or other material to which the Licensor applied this Public
+     License.
+
+  g. Licensed Rights means the rights granted to You subject to the
+     terms and conditions of this Public License, which are limited to
+     all Copyright and Similar Rights that apply to Your use of the
+     Licensed Material and that the Licensor has authority to license.
+
+  h. Licensor means the individual(s) or entity(ies) granting rights
+     under this Public License.
+
+  i. Share means to provide material to the public by any means or
+     process that requires permission under the Licensed Rights, such
+     as reproduction, public display, public performance, distribution,
+     dissemination, communication, or importation, and to make material
+     available to the public including in ways that members of the
+     public may access the material from a place and at a time
+     individually chosen by them.
+
+  j. Sui Generis Database Rights means rights other than copyright
+     resulting from Directive 96/9/EC of the European Parliament and of
+     the Council of 11 March 1996 on the legal protection of databases,
+     as amended and/or succeeded, as well as other essentially
+     equivalent rights anywhere in the world.
+
+  k. You means the individual or entity exercising the Licensed Rights
+     under this Public License. Your has a corresponding meaning.
+
+
+Section 2 -- Scope.
+
+  a. License grant.
+
+       1. Subject to the terms and conditions of this Public License,
+          the Licensor hereby grants You a worldwide, royalty-free,
+          non-sublicensable, non-exclusive, irrevocable license to
+          exercise the Licensed Rights in the Licensed Material to:
+
+            a. reproduce and Share the Licensed Material, in whole or
+               in part; and
+
+            b. produce, reproduce, and Share Adapted Material.
+
+       2. Exceptions and Limitations. For the avoidance of doubt, where
+          Exceptions and Limitations apply to Your use, this Public
+          License does not apply, and You do not need to comply with
+          its terms and conditions.
+
+       3. Term. The term of this Public License is specified in Section
+          6(a).
+
+       4. Media and formats; technical modifications allowed. The
+          Licensor authorizes You to exercise the Licensed Rights in
+          all media and formats whether now known or hereafter created,
+          and to make technical modifications necessary to do so. The
+          Licensor waives and/or agrees not to assert any right or
+          authority to forbid You from making technical modifications
+          necessary to exercise the Licensed Rights, including
+          technical modifications necessary to circumvent Effective
+          Technological Measures. For purposes of this Public License,
+          simply making modifications authorized by this Section 2(a)
+          (4) never produces Adapted Material.
+
+       5. Downstream recipients.
+
+            a. Offer from the Licensor -- Licensed Material. Every
+               recipient of the Licensed Material automatically
+               receives an offer from the Licensor to exercise the
+               Licensed Rights under the terms and conditions of this
+               Public License.
+
+            b. No downstream restrictions. You may not offer or impose
+               any additional or different terms or conditions on, or
+               apply any Effective Technological Measures to, the
+               Licensed Material if doing so restricts exercise of the
+               Licensed Rights by any recipient of the Licensed
+               Material.
+
+       6. No endorsement. Nothing in this Public License constitutes or
+          may be construed as permission to assert or imply that You
+          are, or that Your use of the Licensed Material is, connected
+          with, or sponsored, endorsed, or granted official status by,
+          the Licensor or others designated to receive attribution as
+          provided in Section 3(a)(1)(A)(i).
+
+  b. Other rights.
+
+       1. Moral rights, such as the right of integrity, are not
+          licensed under this Public License, nor are publicity,
+          privacy, and/or other similar personality rights; however, to
+          the extent possible, the Licensor waives and/or agrees not to
+          assert any such rights held by the Licensor to the limited
+          extent necessary to allow You to exercise the Licensed
+          Rights, but not otherwise.
+
+       2. Patent and trademark rights are not licensed under this
+          Public License.
+
+       3. To the extent possible, the Licensor waives any right to
+          collect royalties from You for the exercise of the Licensed
+          Rights, whether directly or through a collecting society
+          under any voluntary or waivable statutory or compulsory
+          licensing scheme. In all other cases the Licensor expressly
+          reserves any right to collect such royalties.
+
+
+Section 3 -- License Conditions.
+
+Your exercise of the Licensed Rights is expressly made subject to the
+following conditions.
+
+  a. Attribution.
+
+       1. If You Share the Licensed Material (including in modified
+          form), You must:
+
+            a. retain the following if it is supplied by the Licensor
+               with the Licensed Material:
+
+                 i. identification of the creator(s) of the Licensed
+                    Material and any others designated to receive
+                    attribution, in any reasonable manner requested by
+                    the Licensor (including by pseudonym if
+                    designated);
+
+                ii. a copyright notice;
+
+               iii. a notice that refers to this Public License;
+
+                iv. a notice that refers to the disclaimer of
+                    warranties;
+
+                 v. a URI or hyperlink to the Licensed Material to the
+                    extent reasonably practicable;
+
+            b. indicate if You modified the Licensed Material and
+               retain an indication of any previous modifications; and
+
+            c. indicate the Licensed Material is licensed under this
+               Public License, and include the text of, or the URI or
+               hyperlink to, this Public License.
+
+       2. You may satisfy the conditions in Section 3(a)(1) in any
+          reasonable manner based on the medium, means, and context in
+          which You Share the Licensed Material. For example, it may be
+          reasonable to satisfy the conditions by providing a URI or
+          hyperlink to a resource that includes the required
+          information.
+
+       3. If requested by the Licensor, You must remove any of the
+          information required by Section 3(a)(1)(A) to the extent
+          reasonably practicable.
+
+       4. If You Share Adapted Material You produce, the Adapter's
+          License You apply must not prevent recipients of the Adapted
+          Material from complying with this Public License.
+
+
+Section 4 -- Sui Generis Database Rights.
+
+Where the Licensed Rights include Sui Generis Database Rights that
+apply to Your use of the Licensed Material:
+
+  a. for the avoidance of doubt, Section 2(a)(1) grants You the right
+     to extract, reuse, reproduce, and Share all or a substantial
+     portion of the contents of the database;
+
+  b. if You include all or a substantial portion of the database
+     contents in a database in which You have Sui Generis Database
+     Rights, then the database in which You have Sui Generis Database
+     Rights (but not its individual contents) is Adapted Material; and
+
+  c. You must comply with the conditions in Section 3(a) if You Share
+     all or a substantial portion of the contents of the database.
+
+For the avoidance of doubt, this Section 4 supplements and does not
+replace Your obligations under this Public License where the Licensed
+Rights include other Copyright and Similar Rights.
+
+
+Section 5 -- Disclaimer of Warranties and Limitation of Liability.
+
+  a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE
+     EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS
+     AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF
+     ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS,
+     IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION,
+     WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR
+     PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS,
+     ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT
+     KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT
+     ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU.
+
+  b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE
+     TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION,
+     NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT,
+     INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES,
+     COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR
+     USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN
+     ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR
+     DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR
+     IN PART, THIS LIMITATION MAY NOT APPLY TO YOU.
+
+  c. The disclaimer of warranties and limitation of liability provided
+     above shall be interpreted in a manner that, to the extent
+     possible, most closely approximates an absolute disclaimer and
+     waiver of all liability.
+
+
+Section 6 -- Term and Termination.
+
+  a. This Public License applies for the term of the Copyright and
+     Similar Rights licensed here. However, if You fail to comply with
+     this Public License, then Your rights under this Public License
+     terminate automatically.
+
+  b. Where Your right to use the Licensed Material has terminated under
+     Section 6(a), it reinstates:
+
+       1. automatically as of the date the violation is cured, provided
+          it is cured within 30 days of Your discovery of the
+          violation; or
+
+       2. upon express reinstatement by the Licensor.
+
+     For the avoidance of doubt, this Section 6(b) does not affect any
+     right the Licensor may have to seek remedies for Your violations
+     of this Public License.
+
+  c. For the avoidance of doubt, the Licensor may also offer the
+     Licensed Material under separate terms or conditions or stop
+     distributing the Licensed Material at any time; however, doing so
+     will not terminate this Public License.
+
+  d. Sections 1, 5, 6, 7, and 8 survive termination of this Public
+     License.
+
+
+Section 7 -- Other Terms and Conditions.
+
+  a. The Licensor shall not be bound by any additional or different
+     terms or conditions communicated by You unless expressly agreed.
+
+  b. Any arrangements, understandings, or agreements regarding the
+     Licensed Material not stated herein are separate from and
+     independent of the terms and conditions of this Public License.
+
+
+Section 8 -- Interpretation.
+
+  a. For the avoidance of doubt, this Public License does not, and
+     shall not be interpreted to, reduce, limit, restrict, or impose
+     conditions on any use of the Licensed Material that could lawfully
+     be made without permission under this Public License.
+
+  b. To the extent possible, if any provision of this Public License is
+     deemed unenforceable, it shall be automatically reformed to the
+     minimum extent necessary to make it enforceable. If the provision
+     cannot be reformed, it shall be severed from this Public License
+     without affecting the enforceability of the remaining terms and
+     conditions.
+
+  c. No term or condition of this Public License will be waived and no
+     failure to comply consented to unless expressly agreed to by the
+     Licensor.
+
+  d. Nothing in this Public License constitutes or may be interpreted
+     as a limitation upon, or waiver of, any privileges and immunities
+     that apply to the Licensor or You, including from the legal
+     processes of any jurisdiction or authority.
+
+
+=======================================================================
+
+Creative Commons is not a party to its public
+licenses. Notwithstanding, Creative Commons may elect to apply one of
+its public licenses to material it publishes and in those instances
+will be considered the “Licensor.” The text of the Creative Commons
+public licenses is dedicated to the public domain under the CC0 Public
+Domain Dedication. Except for the limited purpose of indicating that
+material is shared under a Creative Commons public license or as
+otherwise permitted by the Creative Commons policies published at
+creativecommons.org/policies, Creative Commons does not authorize the
+use of the trademark "Creative Commons" or any other trademark or logo
+of Creative Commons without its prior written consent including,
+without limitation, in connection with any unauthorized modifications
+to any of its public licenses or any other arrangements,
+understandings, or agreements concerning use of licensed material. For
+the avoidance of doubt, this paragraph does not form part of the
+public licenses.
+
+Creative Commons may be contacted at creativecommons.org.

Orient_Anything/README.md

@@ -0,0 +1,124 @@
+<div align="center">
+<h2>Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Models</h2>
+
+[**Zehan Wang**](https://scholar.google.com/citations?user=euXK0lkAAAAJ&hl=zh-CN)<sup>1*</sup> · [**Ziang Zhang**](https://scholar.google.com/citations?hl=zh-CN&user=DptGMnYAAAAJ)<sup>1*</sup> · [**Tianyu Pang**](https://scholar.google.com/citations?hl=zh-CN&user=wYDbtFsAAAAJ)<sup>2</sup> · [**Du Chao**](https://scholar.google.com/citations?hl=zh-CN&user=QOp7xW0AAAAJ)<sup>2</sup> · [**Hengshuang Zhao**](https://scholar.google.com/citations?user=4uE10I0AAAAJ&hl&oi=ao)<sup>3</sup> · [**Zhou Zhao**](https://scholar.google.com/citations?user=IIoFY90AAAAJ&hl&oi=ao)<sup>1</sup>
+
+<sup>1</sup>Zhejiang University&emsp;&emsp;&emsp;&emsp;<sup>2</sup>SEA AI Lab&emsp;&emsp;&emsp;&emsp;<sup>3</sup>HKU
+
+*Equal Contribution
+
+
+<a href='https://arxiv.org/abs/2412.18605'><img src='https://img.shields.io/badge/arXiv-Orient Anything-red' alt='Paper PDF'></a>
+<a href='https://orient-anything.github.io'><img src='https://img.shields.io/badge/Project_Page-Orient Anything-green' alt='Project Page'></a>
+<a href='https://huggingface.co/spaces/Viglong/Orient-Anything'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue'></a>
+<a href='https://huggingface.co/papers/2412.18605'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Paper-yellow'></a>
+</div>
+
+**Orient Anything**, a robust image-based object orientation estimation model. By training on 2M rendered labeled images, it achieves strong zero-shot generalization ability for images in the wild.
+
+![teaser](assets/demo.png)
+
+## News
+* **2025-05-01:** Orient Anything is accepted by ICML 2025!
+* **2024-12-24:** [Paper](https://arxiv.org/abs/2412.18605), [Project Page](https://orient-anything.github.io), [Code](https://github.com/SpatialVision/Orient-Anything), Models, and [Demo](https://huggingface.co/spaces/Viglong/Orient-Anything) are released.
+
+
+
+## Pre-trained models
+
+We provide **three models** of varying scales for robust object orientation estimation in images:
+
+| Model | Params | Checkpoint |
+|:-|-:|:-:|
+| Orient-Anything-Small | 23.3 M | [Download](https://huggingface.co/Viglong/OriNet/blob/main/cropsmallEx03/dino_weight.pt) |
+| Orient-Anything-Base | 87.8 M | [Download](https://huggingface.co/Viglong/OriNet/blob/main/cropbaseEx032/dino_weight.pt) |
+| Orient-Anything-Large | 305 M | [Download](https://huggingface.co/Viglong/OriNet/blob/main/croplargeEX2/dino_weight.pt) |
+
+## Usage
+
+### 1 Prepraration
+
+```bash
+pip install -r requirements.txt
+```
+
+### 2 Use our models
+#### 2.1 In Gradio app
+Start gradio by executing the following script:
+
+```bash
+python app.py
+```
+then open GUI page(default is https://127.0.0.1:7860) in web browser.
+
+or, you can try it in our [Huggingface-Space](https://huggingface.co/spaces/Viglong/Orient-Anything)
+
+#### 2.2 In Python Scripts
+```python
+from paths import *
+from vision_tower import DINOv2_MLP
+from transformers import AutoImageProcessor
+import torch
+from PIL import Image
+
+import torch.nn.functional as F
+from utils import *
+from inference import *
+
+from huggingface_hub import hf_hub_download
+ckpt_path = hf_hub_download(repo_id="Viglong/Orient-Anything", filename="croplargeEX2/dino_weight.pt", repo_type="model", cache_dir='./', resume_download=True)
+print(ckpt_path)
+
+save_path = './'
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+dino = DINOv2_MLP(
+                    dino_mode   = 'large',
+                    in_dim      = 1024,
+                    out_dim     = 360+180+180+2,
+                    evaluate    = True,
+                    mask_dino   = False,
+                    frozen_back = False
+                )
+
+dino.eval()
+print('model create')
+dino.load_state_dict(torch.load(ckpt_path, map_location='cpu'))
+dino = dino.to(device)
+print('weight loaded')
+val_preprocess   = AutoImageProcessor.from_pretrained(DINO_LARGE, cache_dir='./')
+
+image_path = '/path/to/image'
+origin_image = Image.open(image_path).convert('RGB')
+angles = get_3angle(origin_image, dino, val_preprocess, device)
+azimuth     = float(angles[0])
+polar       = float(angles[1])
+rotation    = float(angles[2])
+confidence  = float(angles[3])
+
+
+```
+
+
+### Best Practice
+To avoid ambiguity, our model only supports inputs that contain images of a single object. For daily images that usually contain multiple objects, it is a good choice to isolate each object with DINO-grounding and predict the orientation separately.
+```python
+[ToDo]
+```
+### Test-Time Augmentation
+In order to further enhance the robustness of the model，We further propose the test-time ensemble strategy. The input images will be randomly cropped into different variants, and the predicted orientation of different variants will be voted as the final prediction result. We implement this strategy in functions `get_3angle_infer_aug()` and `get_crop_images()`.
+
+## Citation
+
+If you find this project useful, please consider citing:
+
+```bibtex
+@article{orient_anything,
+  title={Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Models},
+  author={Wang, Zehan and Zhang, Ziang and Pang, Tianyu and Du, Chao and Zhao, Hengshuang and Zhao, Zhou},
+  journal={arXiv:2412.18605},
+  year={2024}
+}
+```
+
+## Acknowledgement
+Thanks to the open source of the following projects: [Grounded-Segment-Anything](https://github.com/IDEA-Research/Grounded-Segment-Anything), [render-py](https://github.com/tvytlx/render-py)

Orient_Anything/Rotation.py

@@ -0,0 +1,97 @@
+from osdsynth.Orient_Anything.paths import *
+from osdsynth.Orient_Anything.vision_tower import DINOv2_MLP
+from transformers import AutoImageProcessor
+import torch
+from PIL import Image
+
+import torch.nn.functional as F
+from osdsynth.Orient_Anything.utils import *
+from osdsynth.Orient_Anything.inference import *
+from osdsynth.utils.logger import SkipImageException
+
+
+class Rotation:
+    def __init__(self, cfg, logger, device):
+        self.rotation_model = DINOv2_MLP(
+                    dino_mode   = 'large',
+                    in_dim      = 1024,
+                    out_dim     = 360+180+180+2,
+                    evaluate    = True,
+                    mask_dino   = False,
+                    frozen_back = False
+                ).eval()
+        self.rotation_model.load_state_dict(torch.load(cfg.rotation_model_ckpt_path, map_location='cpu'))
+        self.rotation_model = self.rotation_model.to(device)
+        self.val_preprocess = AutoImageProcessor.from_pretrained(DINO_LARGE, cache_dir='./')
+        self.logger = logger
+        self.device = device
+        
+    def get_R(self, phi, theta, gamma):
+        R = np.array([[-math.sin(phi), 0, -math.cos(phi)],
+                        [math.cos(phi) * math.sin(theta), -math.cos(theta), -math.sin(phi) * math.sin(theta)],
+                        [-math.cos(phi) * math.cos(theta), -math.sin(theta), math.sin(phi) * math.cos(theta)]
+                    ]) # The position of the object as seen by the camera
+        
+        # R rotates gamma around the z-axis of the world coordinate system
+        gamma = gamma * math.pi / 180
+        R = np.dot(R, np.array([[math.cos(gamma), -math.sin(gamma), 0],
+                                [math.sin(gamma), math.cos(gamma), 0],
+                                [0, 0, 1]
+                                ]))
+        
+        
+        
+        return R
+
+    def process_single(self, image_segment_wbg, do_rm_bkg=True, do_infer_aug=False):
+        origin_img_wbg = Image.fromarray(image_segment_wbg) if isinstance(image_segment_wbg, np.ndarray) else image_segment_wbg
+        if do_infer_aug:
+            rm_bkg_img = background_preprocess(origin_img_wbg, True)
+            angles = get_3angle_infer_aug(origin_img_wbg, rm_bkg_img, self.rotation_model, self.val_preprocess, self.device)
+        else:
+            rm_bkg_img = background_preprocess(origin_img_wbg, False)
+            angles = get_3angle(rm_bkg_img, self.rotation_model, self.val_preprocess, self.device)
+            confidence = float(angles[3])
+            
+            if confidence < 0.5:
+                do_rm_bkg_img = background_preprocess(origin_img_wbg, True)
+                do_angles = get_3angle(do_rm_bkg_img, self.rotation_model, self.val_preprocess, self.device)
+                do_confidence = float(do_angles[3])
+                if do_confidence >0.5 :
+                    rm_bkg_img = do_rm_bkg_img
+                    angles = do_angles
+        
+        
+        phi   = np.radians(angles[0])
+        theta = np.radians(angles[1])
+        gamma = angles[2]
+        confidence = float(angles[3])
+        
+
+
+        render_axis = render_3D_axis(phi, theta, gamma)
+        res_img = overlay_images_with_scaling(render_axis, rm_bkg_img)
+        res_img.save("res_img.png")
+        rotation_matrix = self.get_R(phi=phi, theta=theta, gamma=gamma)
+        return rotation_matrix, confidence
+
+
+    def process(self, detection_list, do_rm_bkg=True, do_infer_aug=False):
+        skip_index = []
+        for i in range(len(detection_list)):
+            # image_segment_wobg = detection_list[i]['image_segment']
+            image_segment_wbg = detection_list[i]['image_crop']
+            
+            # object-centered reference system, the camera's bit position
+            rotation_matrix, confidence = self.process_single(image_segment_wbg, do_rm_bkg=do_rm_bkg, do_infer_aug=do_infer_aug)
+            
+            # For the camera, the position and rotation of the object
+            detection_list[i]['rotation_matrix'] = rotation_matrix 
+            detection_list[i]['rotation_matrix_confidence'] = confidence
+            
+        # 删除不符合条件的项
+        for i in skip_index[::-1]:
+            del detection_list[i]
+        
+            
+        return detection_list

Orient_Anything/app.py

@@ -0,0 +1,72 @@
+import gradio as gr
+from paths import *
+
+from vision_tower import DINOv2_MLP
+from transformers import AutoImageProcessor
+import torch
+from inference import *
+from utils import *
+
+from huggingface_hub import hf_hub_download
+ckpt_path = hf_hub_download(repo_id="Viglong/Orient-Anything", filename="ronormsigma1/dino_weight.pt", repo_type="model", cache_dir='./', resume_download=True)
+print(ckpt_path)
+
+save_path = './'
+device = 'cpu'
+dino = DINOv2_MLP(
+                    dino_mode   = 'large',
+                    in_dim      = 1024,
+                    out_dim     = 360+180+360+2,
+                    evaluate    = True,
+                    mask_dino   = False,
+                    frozen_back = False
+                )
+
+dino.eval()
+print('model create')
+dino.load_state_dict(torch.load(ckpt_path, map_location='cpu'))
+dino = dino.to(device)
+print('weight loaded')
+val_preprocess   = AutoImageProcessor.from_pretrained(DINO_LARGE, cache_dir='./')
+
+def infer_func(img, do_rm_bkg, do_infer_aug):
+    origin_img = Image.fromarray(img)
+    if do_infer_aug:
+        rm_bkg_img = background_preprocess(origin_img, True)
+        angles = get_3angle_infer_aug(origin_img, rm_bkg_img, dino, val_preprocess, device)
+    else:
+        rm_bkg_img = background_preprocess(origin_img, do_rm_bkg)
+        angles = get_3angle(rm_bkg_img, dino, val_preprocess, device)
+    
+    phi   = np.radians(angles[0])
+    theta = np.radians(angles[1])
+    gamma = angles[2]
+    confidence = float(angles[3])
+    if confidence > 0.5:
+        render_axis = render_3D_axis(phi, theta, gamma)
+        res_img = overlay_images_with_scaling(render_axis, rm_bkg_img)
+    else:
+        res_img = img
+    
+    # axis_model = "axis.obj"
+    return [res_img, round(float(angles[0]), 2), round(float(angles[1]), 2), round(float(angles[2]), 2), round(float(angles[3]), 2)]
+
+server = gr.Interface(
+    flagging_mode='never',
+    fn=infer_func, 
+    inputs=[
+        gr.Image(height=512, width=512, label="upload your image"),
+        gr.Checkbox(label="Remove Background", value=True),
+        gr.Checkbox(label="Inference time augmentation", value=False)
+    ], 
+    outputs=[
+        gr.Image(height=512, width=512, label="result image"),
+        # gr.Model3D(clear_color=[0.0, 0.0, 0.0, 0.0],  label="3D Model"),
+        gr.Textbox(lines=1, label='Azimuth(0~360°)'),
+        gr.Textbox(lines=1, label='Polar(-90~90°)'),
+        gr.Textbox(lines=1, label='Rotation(-90~90°)'),
+        gr.Textbox(lines=1, label='Confidence(0~1)')
+    ]
+)
+
+server.launch()

Orient_Anything/inference.py

@@ -0,0 +1,49 @@
+import torch
+from PIL import Image
+from osdsynth.Orient_Anything.utils import *
+import torch.nn.functional as F
+import numpy as np
+
+def get_3angle(image, dino, val_preprocess, device):
+    
+    # image = Image.open(image_path).convert('RGB')
+    image_inputs = val_preprocess(images = image)
+    image_inputs['pixel_values'] = torch.from_numpy(np.array(image_inputs['pixel_values'])).to(device)
+    with torch.no_grad():
+        dino_pred = dino(image_inputs)
+
+    gaus_ax_pred   = torch.argmax(dino_pred[:, 0:360], dim=-1)
+    gaus_pl_pred   = torch.argmax(dino_pred[:, 360:360+180], dim=-1)
+    gaus_ro_pred   = torch.argmax(dino_pred[:, 360+180:360+180+180], dim=-1)
+    confidence     = F.softmax(dino_pred[:, -2:], dim=-1)[0][0]
+    angles = torch.zeros(4)
+    angles[0]  = gaus_ax_pred
+    angles[1]  = gaus_pl_pred - 90
+    angles[2]  = gaus_ro_pred - 90
+    angles[3]  = confidence
+    return angles
+
+def get_3angle_infer_aug(origin_img, rm_bkg_img, dino, val_preprocess, device):
+    
+    # image = Image.open(image_path).convert('RGB')
+    image = get_crop_images(origin_img, num=3) + get_crop_images(rm_bkg_img, num=3)
+    image_inputs = val_preprocess(images = image)
+    image_inputs['pixel_values'] = torch.from_numpy(np.array(image_inputs['pixel_values'])).to(device)
+    with torch.no_grad():
+        dino_pred = dino(image_inputs)
+
+    gaus_ax_pred   = torch.argmax(dino_pred[:, 0:360], dim=-1).to(torch.float32)
+    gaus_pl_pred   = torch.argmax(dino_pred[:, 360:360+180], dim=-1).to(torch.float32)
+    gaus_ro_pred   = torch.argmax(dino_pred[:, 360+180:360+180+180], dim=-1).to(torch.float32)
+    
+    gaus_ax_pred   = remove_outliers_and_average_circular(gaus_ax_pred)
+    gaus_pl_pred   = remove_outliers_and_average(gaus_pl_pred)
+    gaus_ro_pred   = remove_outliers_and_average(gaus_ro_pred)
+    
+    confidence     = torch.mean(F.softmax(dino_pred[:, -2:], dim=-1), dim=0)[0]
+    angles = torch.zeros(4)
+    angles[0]  = gaus_ax_pred
+    angles[1]  = gaus_pl_pred - 90
+    angles[2]  = gaus_ro_pred - 90
+    angles[3]  = confidence
+    return angles

Orient_Anything/paths.py

@@ -0,0 +1,4 @@
+DINO_SMALL  = "facebook/dinov2-small"
+DINO_BASE   = "facebook/dinov2-base"
+DINO_LARGE  = "/mnt/prev_nas/qhy/dinov2-large"
+DINO_GIANT  = "facebook/dinov2-giant"

Orient_Anything/requirements.txt

@@ -0,0 +1,9 @@
+torch==2.2.1
+transformers==4.38
+matplotlib
+pillow==10.2.0
+huggingface-hub==0.26.5
+gradio==5.9.0
+numpy==1.26.4
+onnxruntime
+rembg

Orient_Anything/utils.py

@@ -0,0 +1,304 @@
+import rembg
+import random
+import torch
+import numpy as np
+from PIL import Image, ImageOps
+import PIL
+from typing import Any
+import matplotlib.pyplot as plt
+import io
+
+def resize_foreground(
+    image: Image,
+    ratio: float,
+) -> Image:
+    image = np.array(image)
+    assert image.shape[-1] == 4
+    alpha = np.where(image[..., 3] > 0)
+    y1, y2, x1, x2 = (
+        alpha[0].min(),
+        alpha[0].max(),
+        alpha[1].min(),
+        alpha[1].max(),
+    )
+    # crop the foreground
+    fg = image[y1:y2, x1:x2]
+    # pad to square
+    size = max(fg.shape[0], fg.shape[1])
+    ph0, pw0 = (size - fg.shape[0]) // 2, (size - fg.shape[1]) // 2
+    ph1, pw1 = size - fg.shape[0] - ph0, size - fg.shape[1] - pw0
+    new_image = np.pad(
+        fg,
+        ((ph0, ph1), (pw0, pw1), (0, 0)),
+        mode="constant",
+        constant_values=((0, 0), (0, 0), (0, 0)),
+    )
+
+    # compute padding according to the ratio
+    new_size = int(new_image.shape[0] / ratio)
+    # pad to size, double side
+    ph0, pw0 = (new_size - size) // 2, (new_size - size) // 2
+    ph1, pw1 = new_size - size - ph0, new_size - size - pw0
+    new_image = np.pad(
+        new_image,
+        ((ph0, ph1), (pw0, pw1), (0, 0)),
+        mode="constant",
+        constant_values=((0, 0), (0, 0), (0, 0)),
+    )
+    new_image = Image.fromarray(new_image)
+    return new_image
+
+def remove_background(image: Image,
+    rembg_session: Any = None,
+    force: bool = False,
+    **rembg_kwargs,
+) -> Image:
+    do_remove = True
+    if image.mode == "RGBA" and image.getextrema()[3][0] < 255:
+        do_remove = False
+    do_remove = do_remove or force
+    if do_remove:
+        image = rembg.remove(image, session=rembg_session, **rembg_kwargs)
+    return image
+
+def random_crop(image, crop_scale=(0.8, 0.95)):
+    """
+    随机裁切图片
+        image (numpy.ndarray):  (H, W, C)。
+        crop_scale (tuple): (min_scale, max_scale)。
+    """
+    assert isinstance(image, Image.Image), "iput must be PIL.Image.Image"
+    assert len(crop_scale) == 2 and 0 < crop_scale[0] <= crop_scale[1] <= 1
+
+    width, height = image.size
+
+    # 计算裁切的高度和宽度
+    crop_width = random.randint(int(width * crop_scale[0]), int(width * crop_scale[1]))
+    crop_height = random.randint(int(height * crop_scale[0]), int(height * crop_scale[1]))
+
+    # 随机选择裁切的起始点
+    left = random.randint(0, width - crop_width)
+    top = random.randint(0, height - crop_height)
+
+    # 裁切图片
+    cropped_image = image.crop((left, top, left + crop_width, top + crop_height))
+
+    return cropped_image
+
+def get_crop_images(img, num=3):
+    cropped_images = []
+    for i in range(num):
+        cropped_images.append(random_crop(img))
+    return cropped_images
+
+def background_preprocess(input_image, do_remove_background):
+
+    rembg_session = rembg.new_session() if do_remove_background else None
+
+    if do_remove_background:
+        input_image = remove_background(input_image, rembg_session)
+        input_image = resize_foreground(input_image, 0.85)
+
+    return input_image
+
+def remove_outliers_and_average(tensor, threshold=1.5):
+    assert tensor.dim() == 1, "dimension of input Tensor must equal to 1"
+
+    q1 = torch.quantile(tensor, 0.25)
+    q3 = torch.quantile(tensor, 0.75)
+    iqr = q3 - q1
+
+    lower_bound = q1 - threshold * iqr
+    upper_bound = q3 + threshold * iqr
+
+    non_outliers = tensor[(tensor >= lower_bound) & (tensor <= upper_bound)]
+
+    if len(non_outliers) == 0:
+        return tensor.mean().item()
+
+    return non_outliers.mean().item()
+
+
+def remove_outliers_and_average_circular(tensor, threshold=1.5):
+    assert tensor.dim() == 1, "dimension of input Tensor must equal to 1"
+
+    # 将角度转换为二维平面上的点
+    radians = tensor * torch.pi / 180.0
+    x_coords = torch.cos(radians)
+    y_coords = torch.sin(radians)
+
+    # 计算平均向量
+    mean_x = torch.mean(x_coords)
+    mean_y = torch.mean(y_coords)
+
+    differences = torch.sqrt((x_coords - mean_x) * (x_coords - mean_x) + (y_coords - mean_y) * (y_coords - mean_y))
+
+    # 计算四分位数和 IQR
+    q1 = torch.quantile(differences, 0.25)
+    q3 = torch.quantile(differences, 0.75)
+    iqr = q3 - q1
+
+    # 计算上下限
+    lower_bound = q1 - threshold * iqr
+    upper_bound = q3 + threshold * iqr
+
+    # 筛选非离群点
+    non_outliers = tensor[(differences >= lower_bound) & (differences <= upper_bound)]
+
+    if len(non_outliers) == 0:
+        mean_angle = torch.atan2(mean_y, mean_x) * 180.0 / torch.pi
+        mean_angle = (mean_angle + 360) % 360
+        return mean_angle  # 如果没有非离群点，返回 None
+
+    # 对非离群点再次计算平均向量
+    radians = non_outliers * torch.pi / 180.0
+    x_coords = torch.cos(radians)
+    y_coords = torch.sin(radians)
+
+    mean_x = torch.mean(x_coords)
+    mean_y = torch.mean(y_coords)
+
+    mean_angle = torch.atan2(mean_y, mean_x) * 180.0 / torch.pi
+    mean_angle = (mean_angle + 360) % 360
+
+    return mean_angle
+
+def scale(x):
+    # print(x)
+    # if abs(x[0])<0.1 and abs(x[1])<0.1:
+        
+    #     return x*5
+    # else:
+    #     return x
+    return x*3
+
+def get_proj2D_XYZ(phi, theta, gamma):
+    x = np.array([-1*np.sin(phi)*np.cos(gamma) - np.cos(phi)*np.sin(theta)*np.sin(gamma), np.sin(phi)*np.sin(gamma) - np.cos(phi)*np.sin(theta)*np.cos(gamma)])
+    y = np.array([-1*np.cos(phi)*np.cos(gamma) + np.sin(phi)*np.sin(theta)*np.sin(gamma), np.cos(phi)*np.sin(gamma) + np.sin(phi)*np.sin(theta)*np.cos(gamma)])
+    z = np.array([np.cos(theta)*np.sin(gamma), np.cos(theta)*np.cos(gamma)])
+    x = scale(x)
+    y = scale(y)
+    z = scale(z)
+    return x, y, z
+
+# 绘制3D坐标轴
+def draw_axis(ax, origin, vector, color, label=None):
+    ax.quiver(origin[0], origin[1], vector[0], vector[1], angles='xy', scale_units='xy', scale=1, color=color)
+    if label!=None:
+        ax.text(origin[0] + vector[0] * 1.1, origin[1] + vector[1] * 1.1, label, color=color, fontsize=12)
+
+def matplotlib_2D_arrow(angles, rm_bkg_img):
+    fig, ax = plt.subplots(figsize=(8, 8))
+
+    # 设置旋转角度
+    phi   = np.radians(angles[0])
+    theta = np.radians(angles[1])
+    gamma = np.radians(-1*angles[2])
+
+    w, h = rm_bkg_img.size
+    if h>w:
+        extent = [-5*w/h, 5*w/h, -5, 5]
+    else:
+        extent = [-5, 5, -5*h/w, 5*h/w]
+    ax.imshow(rm_bkg_img, extent=extent, zorder=0, aspect ='auto')  # extent 设置图片的显示范围
+
+    origin = np.array([0, 0])
+
+    # 旋转后的向量
+    rot_x, rot_y, rot_z = get_proj2D_XYZ(phi, theta, gamma)
+
+    # draw arrow
+    arrow_attr = [{'point':rot_x, 'color':'r', 'label':'front'}, 
+                  {'point':rot_y, 'color':'g', 'label':'right'}, 
+                  {'point':rot_z, 'color':'b', 'label':'top'}]
+    
+    if phi> 45 and phi<=225:
+        order = [0,1,2]
+    elif phi > 225 and phi < 315:
+        order = [2,0,1]
+    else:
+        order = [2,1,0]
+    
+    for i in range(3):
+        draw_axis(ax, origin, arrow_attr[order[i]]['point'], arrow_attr[order[i]]['color'], arrow_attr[order[i]]['label'])
+        # draw_axis(ax, origin, rot_y, 'g', label='right')
+        # draw_axis(ax, origin, rot_z, 'b', label='top')
+        # draw_axis(ax, origin, rot_x, 'r', label='front')
+
+    # 关闭坐标轴和网格
+    ax.set_axis_off()
+    ax.grid(False)
+
+    # 设置坐标范围
+    ax.set_xlim(-5, 5)
+    ax.set_ylim(-5, 5)
+
+def figure_to_img(fig):
+    with io.BytesIO() as buf:
+        fig.savefig(buf, format='JPG', bbox_inches='tight')
+        buf.seek(0)
+        image = Image.open(buf).copy()
+    return image
+
+from render import render, Model
+import math
+axis_model = Model("/mnt/prev_nas/qhy_1/GenSpace/osdsynth/Orient_Anything/assets/axis.obj", texture_filename="/mnt/prev_nas/qhy_1/GenSpace/osdsynth/Orient_Anything/assets/axis.png")
+def render_3D_axis(phi, theta, gamma):
+    radius = 240
+    # camera_location = [radius * math.cos(phi), radius * math.sin(phi), radius * math.tan(theta)]
+    # print(camera_location)
+    camera_location = [-1*radius * math.cos(phi), -1*radius * math.tan(theta), radius * math.sin(phi)]
+    img = render(
+        # Model("res/jinx.obj", texture_filename="res/jinx.tga"),
+        axis_model,
+        height=512,
+        width=512,
+        filename="tmp_render.png",
+        cam_loc = camera_location
+    )
+    img = img.rotate(gamma)
+    return img
+
+def overlay_images_with_scaling(center_image: Image.Image, background_image, target_size=(512, 512)):
+    """
+    调整前景图像大小为 512x512，将背景图像缩放以适配，并中心对齐叠加
+    :param center_image: 前景图像
+    :param background_image: 背景图像
+    :param target_size: 前景图像的目标大小，默认 (512, 512)
+    :return: 叠加后的图像
+    """
+    # 确保输入图像为 RGBA 模式
+    if center_image.mode != "RGBA":
+        center_image = center_image.convert("RGBA")
+    if background_image.mode != "RGBA":
+        background_image = background_image.convert("RGBA")
+    
+    # 调整前景图像大小
+    center_image = center_image.resize(target_size)
+    
+    # 缩放背景图像，确保其适合前景图像的尺寸
+    bg_width, bg_height = background_image.size
+    
+    # 按宽度或高度等比例缩放背景
+    scale = target_size[0] / max(bg_width, bg_height)
+    new_width = int(bg_width * scale)
+    new_height = int(bg_height * scale)
+    resized_background = background_image.resize((new_width, new_height))
+    # 计算需要的填充量
+    pad_width = target_size[0] - new_width
+    pad_height = target_size[0] - new_height
+
+    # 计算上下左右的 padding
+    left = pad_width // 2
+    right = pad_width - left
+    top = pad_height // 2
+    bottom = pad_height - top
+
+    # 添加 padding
+    resized_background = ImageOps.expand(resized_background, border=(left, top, right, bottom), fill=(255,255,255,255))
+    
+    # 将前景图像叠加到背景图像上
+    result = resized_background.copy()
+    result.paste(center_image, (0, 0), mask=center_image)
+    
+    return result

Orient_Anything/vision_tower.py

@@ -0,0 +1,164 @@
+import sys
+sys.path.append("/mnt/prev_nas/qhy_1/GenSpace/osdsynth/Orient_Anything")
+
+import torch
+from torch import nn
+import torch.nn.init as init
+import torch.nn.functional as F
+
+from paths import *
+
+from typing import Dict, List, Optional, Set, Tuple, Union
+from transformers import AutoImageProcessor, AutoModel, Dinov2Model
+from transformers.models.dinov2.modeling_dinov2 import Dinov2Embeddings
+from transformers.models.dinov2.configuration_dinov2 import Dinov2Config
+import numpy as np
+from contextlib import nullcontext
+
+def get_activation(activation):
+    if activation.lower() == 'gelu':
+        return nn.GELU()
+    elif activation.lower() == 'rrelu':
+        return nn.RReLU(inplace=True)
+    elif activation.lower() == 'selu':
+        return nn.SELU(inplace=True)
+    elif activation.lower() == 'silu':
+        return nn.SiLU(inplace=True)
+    elif activation.lower() == 'hardswish':
+        return nn.Hardswish(inplace=True)
+    elif activation.lower() == 'leakyrelu':
+        return nn.LeakyReLU(inplace=True)
+    elif activation.lower() == 'sigmoid':
+        return nn.Sigmoid()
+    elif activation.lower() == 'tanh':
+        return nn.Tanh()
+    else:
+        return nn.ReLU(inplace=True)
+
+
+
+class MLP_dim(nn.Module):
+    def __init__(
+        self, in_dim=512, out_dim=1024, bias=True, activation='relu'):
+        super().__init__()
+        self.act = get_activation(activation)
+        self.net1 = nn.Sequential(
+            nn.Linear(in_dim, int(out_dim), bias=bias),
+            nn.BatchNorm1d(int(out_dim)),
+            self.act
+        )
+        self.net2 = nn.Sequential(
+            nn.Linear(int(out_dim), out_dim, bias=bias),
+            nn.BatchNorm1d(out_dim)
+        )
+
+    def forward(self, x):
+        return self.net2(self.net1(x))
+
+class FLIP_Dinov2Embeddings(Dinov2Embeddings):
+    """
+    Construct the CLS token, mask token, position and patch embeddings.
+    """
+
+    def __init__(self, config: Dinov2Config) -> None:
+        super().__init__(config)
+
+    def forward(self, pixel_values: torch.Tensor, bool_masked_pos: Optional[torch.Tensor] = None) -> torch.Tensor:
+        batch_size, _, height, width = pixel_values.shape
+        target_dtype = self.patch_embeddings.projection.weight.dtype
+        embeddings = self.patch_embeddings(pixel_values.to(dtype=target_dtype))
+
+        # add the [CLS] token to the embedded patch tokens
+        cls_tokens = self.cls_token.expand(batch_size, -1, -1)
+        embeddings = torch.cat((cls_tokens, embeddings), dim=1)
+
+        # add positional encoding to each token
+        embeddings = embeddings + self.interpolate_pos_encoding(embeddings, height, width)
+
+        if bool_masked_pos is not None:
+            # embeddings = torch.where(
+            #     bool_masked_pos.unsqueeze(-1), self.mask_token.to(embeddings.dtype).unsqueeze(0), embeddings
+            # )
+            B,S,D = embeddings.shape
+            batch_indices = torch.arange(B).unsqueeze(1)
+            embeddings = embeddings[batch_indices, bool_masked_pos]
+
+        embeddings = self.dropout(embeddings)
+
+        return embeddings
+
+class FLIP_DINOv2(Dinov2Model):
+    def __init__(self, config):
+        super().__init__(config)
+        
+        self.embeddings = FLIP_Dinov2Embeddings(config)
+        
+class DINOv2_MLP(nn.Module):
+    def __init__(self, 
+                 dino_mode,
+                 in_dim,
+                 out_dim,
+                 evaluate,
+                 mask_dino,
+                 frozen_back
+                ) -> None:
+        super().__init__()
+        # self.dinov2 = AutoModel.from_pretrained(DINO_BASE)
+        if dino_mode == 'base':
+            self.dinov2 = FLIP_DINOv2.from_pretrained(DINO_BASE, cache_dir='./')
+        elif dino_mode == 'large':
+            self.dinov2 = FLIP_DINOv2.from_pretrained(DINO_LARGE, cache_dir='./')
+        elif dino_mode == 'small':
+            self.dinov2 = FLIP_DINOv2.from_pretrained(DINO_SMALL, cache_dir='./')
+        elif dino_mode == 'giant':
+            self.dinov2 = FLIP_DINOv2.from_pretrained(DINO_GIANT, cache_dir='./')
+        
+        self.down_sampler = MLP_dim(in_dim=in_dim, out_dim=out_dim)
+        self.random_mask  = False
+        if not evaluate:
+            self.init_weights(self.down_sampler)
+            self.random_mask = mask_dino
+        if frozen_back:
+            self.forward_mode = torch.no_grad()
+        else:
+            self.forward_mode = nullcontext()
+        
+    def forward(self, img_inputs):
+        device = self.get_device()
+        # print(img_inputs['pixel_values'].shape)
+        
+        with self.forward_mode:
+            if self.random_mask:
+                B = len(img_inputs['pixel_values'])
+                S = 256
+                indices = []
+                for i in range(B):
+                    tmp = torch.randperm(S)[:S//2]
+                    tmp = tmp.sort().values + 1
+                    indices.append(tmp)
+                indices = torch.stack(indices, dim=0)
+                indices = torch.cat([torch.zeros(B, 1, dtype=torch.long, device='cpu'), indices], dim=1)
+                # print(indices.shape)
+                img_inputs['bool_masked_pos'] = indices.to(device)
+                
+            dino_outputs = self.dinov2(**img_inputs)
+            dino_seq = dino_outputs.last_hidden_state
+            # B,S,_ = dino_seq.shape
+            # dino_seq = dino_seq.view(B*S,-1)
+            dino_seq = dino_seq[:,0,:]
+        
+        down_sample_out = self.down_sampler(dino_seq)
+        # down_sample_out = down_sample_out.view(B,S,-1)
+        # down_sample_out = down_sample_out[:,0,:]
+        
+        return down_sample_out
+    
+    def get_device(self):
+        return next(self.parameters()).device
+    
+    def init_weights(self, m):
+        if isinstance(m, nn.Linear):
+            init.xavier_uniform_(m.weight)
+            if m.bias is not None:
+                init.constant_(m.bias, 0)
+

external/Grounded-Segment-Anything/.gitmodules

@@ -0,0 +1,7 @@
+
+[submodule "grounded-sam-osx"]
+	path = grounded-sam-osx
+	url = https://github.com/linjing7/grounded-sam-osx.git
+[submodule "VISAM"]
+	path = VISAM
+	url = https://github.com/BingfengYan/VISAM

external/Grounded-Segment-Anything/CITATION.cff

@@ -0,0 +1,8 @@
+cff-version: 1.2.0
+message: "If you use this software, please cite it as below."
+authors:
+  - name: "Grounded-SAM Contributors"
+title: "Grounded-Segment-Anything"
+date-released: 2023-04-06
+url: "https://github.com/IDEA-Research/Grounded-Segment-Anything"
+license: Apache-2.0

external/Grounded-Segment-Anything/Dockerfile

@@ -0,0 +1,30 @@
+FROM pytorch/pytorch:1.13.1-cuda11.6-cudnn8-devel
+
+# Arguments to build Docker Image using CUDA
+ARG USE_CUDA=0
+ARG TORCH_ARCH=
+
+ENV AM_I_DOCKER True
+ENV BUILD_WITH_CUDA "${USE_CUDA}"
+ENV TORCH_CUDA_ARCH_LIST "${TORCH_ARCH}"
+ENV CUDA_HOME /usr/local/cuda-11.6/
+
+RUN mkdir -p /home/appuser/Grounded-Segment-Anything
+COPY . /home/appuser/Grounded-Segment-Anything/
+
+RUN apt-get update && apt-get install --no-install-recommends wget ffmpeg=7:* \
+    libsm6=2:* libxext6=2:* git=1:* nano=2.* \
+    vim=2:* -y \
+    && apt-get clean && apt-get autoremove && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /home/appuser/Grounded-Segment-Anything
+RUN python -m pip install --no-cache-dir -e segment_anything
+
+# When using build isolation, PyTorch with newer CUDA is installed and can't compile GroundingDINO
+RUN python -m pip install --no-cache-dir wheel
+RUN python -m pip install --no-cache-dir --no-build-isolation -e GroundingDINO
+
+WORKDIR /home/appuser
+RUN pip install --no-cache-dir diffusers[torch]==0.15.1 opencv-python==4.7.0.72 \
+    pycocotools==2.0.6 matplotlib==3.5.3 \
+    onnxruntime==1.14.1 onnx==1.13.1 ipykernel==6.16.2 scipy gradio openai

external/Grounded-Segment-Anything/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2020 - present, IDEA, Inc
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

external/Grounded-Segment-Anything/Makefile

@@ -0,0 +1,43 @@
+# Get version of CUDA and enable it for compilation if CUDA > 11.0
+# This solves https://github.com/IDEA-Research/Grounded-Segment-Anything/issues/53
+# and https://github.com/IDEA-Research/Grounded-Segment-Anything/issues/84
+# when running in Docker
+# Check if nvcc is installed
+NVCC := $(shell which nvcc)
+ifeq ($(NVCC),)
+	# NVCC not found
+	USE_CUDA := 0
+	NVCC_VERSION := "not installed"
+else
+	NVCC_VERSION := $(shell nvcc --version | grep -oP 'release \K[0-9.]+')
+	USE_CUDA := $(shell echo "$(NVCC_VERSION) > 11" | bc -l)
+endif
+
+# Add the list of supported ARCHs
+ifeq ($(USE_CUDA), 1)
+	TORCH_CUDA_ARCH_LIST := "3.5;5.0;6.0;6.1;7.0;7.5;8.0;8.6+PTX"
+	BUILD_MESSAGE := "I will try to build the image with CUDA support"
+else
+	TORCH_CUDA_ARCH_LIST :=
+	BUILD_MESSAGE := "CUDA $(NVCC_VERSION) is not supported"
+endif
+
+
+build-image:
+	@echo $(BUILD_MESSAGE)
+	docker build --build-arg USE_CUDA=$(USE_CUDA) \
+	--build-arg TORCH_ARCH=$(TORCH_CUDA_ARCH_LIST) \
+	-t gsa:v0 .
+run:
+ifeq (,$(wildcard ./sam_vit_h_4b8939.pth))
+	wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth
+endif
+ifeq (,$(wildcard ./groundingdino_swint_ogc.pth))
+	wget https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth
+endif
+	docker run --gpus all -it --rm --net=host --privileged \
+	-v /tmp/.X11-unix:/tmp/.X11-unix \
+	-v "${PWD}":/home/appuser/Grounded-Segment-Anything \
+	-e DISPLAY=$DISPLAY \
+	--name=gsa \
+	--ipc=host -it gsa:v0

external/Grounded-Segment-Anything/README.md

@@ -0,0 +1,808 @@
+![](./assets/Grounded-SAM_logo.png)
+
+# Grounded-Segment-Anything
+[![YouTube](https://badges.aleen42.com/src/youtube.svg)](https://youtu.be/oEQYStnF2l8) [![Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/automated-dataset-annotation-and-evaluation-with-grounding-dino-and-sam.ipynb) [![Open in Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://github.com/camenduru/grounded-segment-anything-colab) [![HuggingFace Space](https://img.shields.io/badge/🤗-HuggingFace%20Space-cyan.svg)](https://huggingface.co/spaces/IDEA-Research/Grounded-SAM) [![Replicate](https://replicate.com/cjwbw/grounded-recognize-anything/badge)](https://replicate.com/cjwbw/grounded-recognize-anything)  [![ModelScope Official Demo](https://img.shields.io/badge/ModelScope-Official%20Demo-important)](https://modelscope.cn/studios/tuofeilunhifi/Grounded-Segment-Anything/summary) [![Huggingface Demo by Community](https://img.shields.io/badge/Huggingface-Demo%20by%20Community-red)](https://huggingface.co/spaces/yizhangliu/Grounded-Segment-Anything) [![Stable-Diffusion WebUI](https://img.shields.io/badge/Stable--Diffusion-WebUI%20by%20Community-critical)](https://github.com/continue-revolution/sd-webui-segment-anything) [![Jupyter Notebook Demo](https://img.shields.io/badge/Demo-Jupyter%20Notebook-informational)](./grounded_sam.ipynb) [![Static Badge](https://img.shields.io/badge/GroundingDINO-arXiv-blue)](https://arxiv.org/abs/2303.05499) [![Static Badge](https://img.shields.io/badge/Segment_Anything-arXiv-blue)](https://arxiv.org/abs/2304.02643) [![Static Badge](https://img.shields.io/badge/Grounded_SAM-arXiv-blue)](https://arxiv.org/abs/2401.14159)
+
+
+We plan to create a very interesting demo by combining [Grounding DINO](https://github.com/IDEA-Research/GroundingDINO) and [Segment Anything](https://github.com/facebookresearch/segment-anything) which aims to detect and segment anything with text inputs! And we will continue to improve it and create more interesting demos based on this foundation. And we have already released an overall technical report about our project on arXiv, please check [Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks](https://arxiv.org/abs/2401.14159) for more details.
+
+- 🔥 **[Grounded SAM 2](https://github.com/IDEA-Research/Grounded-SAM-2)** is released now, which combines Grounding DINO with [SAM 2](https://github.com/facebookresearch/segment-anything-2) for any object tracking in open-world scenarios.
+- 🔥 **[Grounding DINO 1.5](https://github.com/IDEA-Research/Grounding-DINO-1.5-API)** is released now, which is IDEA Research's **Most Capable** Open-World Object Detection Model!
+- 🔥 **[Grounding DINO](https://arxiv.org/abs/2303.05499)** and **[Grounded SAM](https://arxiv.org/abs/2401.14159)** are now supported in Huggingface. For more convenient use, you can refer to [this documentation](https://huggingface.co/docs/transformers/model_doc/grounding-dino)
+
+We are very willing to **help everyone share and promote new projects** based on Segment-Anything, Please check out here for more amazing demos and works in the community: [Highlight Extension Projects](#highlighted-projects). You can submit a new issue (with `project` tag) or a new pull request to add new project's links. 
+
+![](./assets/grounded_sam_new_demo_image.png)
+
+![](./assets/ram_grounded_sam_new.png)
+
+**🍄 Why Building this Project?**
+
+The **core idea** behind this project is to **combine the strengths of different models in order to build a very powerful pipeline for solving complex problems**. And it's worth mentioning that this is a workflow for combining strong expert models, where **all parts can be used separately or in combination, and can be replaced with any similar but different models (like replacing Grounding DINO with GLIP or other detectors / replacing Stable-Diffusion with ControlNet or GLIGEN/ Combining with ChatGPT)**.
+
+**🍇 Updates**
+- **`2024/01/26`** We have released a comprehensive technical report about our project on arXiv, please check [Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks](https://arxiv.org/abs/2401.14159) for more details. And we are profoundly grateful for the contributions of all the contributors in this project.
+- **`2023/12/17`** Support [Grounded-RepViT-SAM](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-repvit-sam-demo) demo, thanks a lot for their great work!
+- **`2023/12/16`** Support [Grounded-Edge-SAM](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-edge-sam-demo) demo, thanks a lot for their great work!
+- **`2023/12/10`** Support [Grounded-Efficient-SAM](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-efficient-sam-demo) demo, thanks a lot for their great work!
+- **`2023/11/24`** Release [RAM++](https://arxiv.org/abs/2310.15200), which is the next generation of RAM. RAM++ can recognize any category with high accuracy, including both predefined common categories and diverse open-set categories.
+- **`2023/11/23`** Release our newly proposed visual prompt counting model [T-Rex](https://github.com/IDEA-Research/T-Rex). The introduction [Video](https://www.youtube.com/watch?v=engIEhZogAQ) and [Demo](https://deepdataspace.com/playground/ivp) is available in [DDS](https://github.com/IDEA-Research/deepdataspace) now.
+- **`2023/07/25`** Support [Light-HQ-SAM](https://github.com/SysCV/sam-hq) in [EfficientSAM](./EfficientSAM/), credits to [Mingqiao Ye](https://github.com/ymq2017) and [Lei Ke](https://github.com/lkeab), thanks a lot for their great work!
+- **`2023/07/14`** Combining **Grounding-DINO-B** with [SAM-HQ](https://github.com/SysCV/sam-hq) achieves **49.6 mean AP** in [Segmentation in the Wild](https://eval.ai/web/challenges/challenge-page/1931/overview) competition zero-shot track, surpassing Grounded-SAM by **3.6 mean AP**, thanks for their great work!
+- **`2023/06/28`** Combining Grounding-DINO with Efficient SAM variants including [FastSAM](https://github.com/CASIA-IVA-Lab/FastSAM) and [MobileSAM](https://github.com/ChaoningZhang/MobileSAM) in [EfficientSAM](./EfficientSAM/) for faster annotating, thanks a lot for their great work!
+- **`2023/06/20`** By combining **Grounding-DINO-L** with **SAM-ViT-H**, Grounded-SAM achieves 46.0 mean AP in [Segmentation in the Wild](https://eval.ai/web/challenges/challenge-page/1931/overview) competition zero-shot track on [CVPR 2023 workshop](https://computer-vision-in-the-wild.github.io/cvpr-2023/), surpassing [UNINEXT (CVPR 2023)](https://github.com/MasterBin-IIAU/UNINEXT) by about **4 mean AP**.
+- **`2023/06/16`** Release [RAM-Grounded-SAM Replicate Online Demo](https://replicate.com/cjwbw/ram-grounded-sam). Thanks a lot to [Chenxi](https://chenxwh.github.io/) for providing this nice demo 🌹.
+- **`2023/06/14`** Support [RAM-Grounded-SAM & SAM-HQ](./automatic_label_ram_demo.py) and update [Simple Automatic Label Demo](./automatic_label_ram_demo.py) to support [RAM](https://github.com/OPPOMKLab/recognize-anything), setting up a strong automatic annotation pipeline.
+- **`2023/06/13`** Checkout the [Autodistill: Train YOLOv8 with ZERO Annotations](https://youtu.be/gKTYMfwPo4M) tutorial to learn how to use Grounded-SAM + [Autodistill](https://github.com/autodistill/autodistill) for automated data labeling and real-time model training.
+- **`2023/06/13`** Support [SAM-HQ](https://github.com/SysCV/sam-hq) in [Grounded-SAM Demo](#running_man-grounded-sam-detect-and-segment-everything-with-text-prompt) for higher quality prediction.
+- **`2023/06/12`** Support [RAM-Grounded-SAM](#label-grounded-sam-with-ram-or-tag2text-for-automatic-labeling) for strong automatic labeling pipeline! Thanks for [Recognize-Anything](https://github.com/OPPOMKLab/recognize-anything).
+- **`2023/06/01`** Our Grounded-SAM has been accepted to present a **demo** at [ICCV 2023](https://iccv2023.thecvf.com/)! See you in Paris!
+- **`2023/05/23`**: Support `Image-Referring-Segment`, `Audio-Referring-Segment` and `Text-Referring-Segment` in [ImageBind-SAM](./playground/ImageBind_SAM/).
+- **`2023/05/03`**: Checkout the [Automated Dataset Annotation and Evaluation with GroundingDINO and SAM](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/automated-dataset-annotation-and-evaluation-with-grounding-dino-and-sam.ipynb) which is an amazing tutorial on automatic labeling! Thanks a lot for [Piotr Skalski](https://github.com/SkalskiP) and [Roboflow](https://github.com/roboflow/notebooks)!
+
+
+## Table of Contents
+- [Grounded-Segment-Anything](#grounded-segment-anything)
+  - [Preliminary Works](#preliminary-works)
+  - [Highlighted Projects](#highlighted-projects)
+- [Installation](#installation)
+  - [Install with Docker](#install-with-docker)
+  - [Install locally](#install-without-docker)
+- [Grounded-SAM Playground](#grounded-sam-playground)
+  - [Step-by-Step Notebook Demo](#open_book-step-by-step-notebook-demo)
+  - [GroundingDINO: Detect Everything with Text Prompt](#running_man-groundingdino-detect-everything-with-text-prompt)
+  - [Grounded-SAM: Detect and Segment Everything with Text Prompt](#running_man-grounded-sam-detect-and-segment-everything-with-text-prompt)
+  - [Grounded-SAM with Inpainting: Detect, Segment and Generate Everything with Text Prompt](#skier-grounded-sam-with-inpainting-detect-segment-and-generate-everything-with-text-prompt)
+  - [Grounded-SAM and Inpaint Gradio APP](#golfing-grounded-sam-and-inpaint-gradio-app)
+  - [Grounded-SAM with RAM or Tag2Text for Automatic Labeling](#label-grounded-sam-with-ram-or-tag2text-for-automatic-labeling)
+  - [Grounded-SAM with BLIP & ChatGPT for Automatic Labeling](#robot-grounded-sam-with-blip-for-automatic-labeling)
+  - [Grounded-SAM with Whisper: Detect and Segment Anything with Audio](#open_mouth-grounded-sam-with-whisper-detect-and-segment-anything-with-audio)
+  - [Grounded-SAM ChatBot with Visual ChatGPT](#speech_balloon-grounded-sam-chatbot-demo)
+  - [Grounded-SAM with OSX for 3D Whole-Body Mesh Recovery](#man_dancing-run-grounded-segment-anything--osx-demo)
+  - [Grounded-SAM with VISAM for Tracking and Segment Anything](#man_dancing-run-grounded-segment-anything--visam-demo)
+  - [Interactive Fashion-Edit Playground: Click for Segmentation And Editing](#dancers-interactive-editing)
+  - [Interactive Human-face Editing Playground: Click And Editing Human Face](#dancers-interactive-editing)
+  - [3D Box Via Segment Anything](#camera-3d-box-via-segment-anything)
+  - [Playground: More Interesting and Imaginative Demos with Grounded-SAM](./playground/)
+    - [DeepFloyd: Image Generation with Text Prompt](./playground/DeepFloyd/)
+    - [PaintByExample: Exemplar-based Image Editing with Diffusion Models](./playground/PaintByExample/)
+    - [LaMa: Resolution-robust Large Mask Inpainting with Fourier Convolutions](./playground/LaMa/)
+    - [RePaint: Inpainting using Denoising Diffusion Probabilistic Models](./playground/RePaint/)
+    - [ImageBind with SAM: Segment with Different Modalities](./playground/ImageBind_SAM/)
+  - [Efficient SAM Series for Faster Annotation](./EfficientSAM/)
+    - [Grounded-FastSAM Demo](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-fastsam-demo)
+    - [Grounded-MobileSAM Demo](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-mobilesam-demo)
+    - [Grounded-Light-HQSAM Demo](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-light-hqsam-demo)
+    - [Grounded-Efficient-SAM Demo](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-efficient-sam-demo)
+    - [Grounded-Edge-SAM Demo](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-edge-sam-demo)
+    - [Grounded-RepViT-SAM Demo](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/main/EfficientSAM#run-grounded-repvit-sam-demo)
+- [Citation](#citation)
+
+## Preliminary Works
+
+Here we provide some background knowledge that you may need to know before trying the demos.
+
+<div align="center">
+
+| Title | Intro | Description | Links |
+|:----:|:----:|:----:|:----:|
+| [Segment-Anything](https://arxiv.org/abs/2304.02643) | ![](https://github.com/facebookresearch/segment-anything/blob/main/assets/model_diagram.png?raw=true) | A strong foundation model aims to segment everything in an image, which needs prompts (as boxes/points/text) to generate masks | [[Github](https://github.com/facebookresearch/segment-anything)] <br> [[Page](https://segment-anything.com/)] <br> [[Demo](https://segment-anything.com/demo)] |
+| [Grounding DINO](https://arxiv.org/abs/2303.05499) | ![](https://github.com/IDEA-Research/GroundingDINO/blob/main/.asset/hero_figure.png?raw=True) | A strong zero-shot detector which is capable of to generate high quality boxes and labels with free-form text. | [[Github](https://github.com/IDEA-Research/GroundingDINO)] <br> [[Demo](https://huggingface.co/spaces/ShilongLiu/Grounding_DINO_demo)] |
+| [OSX](http://arxiv.org/abs/2303.16160) | ![](https://github.com/IDEA-Research/OSX/blob/main/assets/demo_video.gif?raw=True) | A strong and efficient one-stage motion capture method to generate high quality 3D human mesh from monucular image. OSX also releases a large-scale upper-body dataset UBody for a more accurate reconstrution in the upper-body scene. | [[Github](https://github.com/IDEA-Research/OSX)] <br> [[Page](https://osx-ubody.github.io/)] <br> [[Video](https://osx-ubody.github.io/)] <br> [[Data](https://docs.google.com/forms/d/e/1FAIpQLSehgBP7wdn_XznGAM2AiJPiPLTqXXHw5uX9l7qeQ1Dh9HoO_A/viewform)] |
+| [Stable-Diffusion](https://arxiv.org/abs/2112.10752) | ![](https://github.com/CompVis/stable-diffusion/blob/main/assets/stable-samples/txt2img/merged-0006.png?raw=True) | A super powerful open-source latent text-to-image diffusion model | [[Github](https://github.com/CompVis/stable-diffusion)] <br> [[Page](https://ommer-lab.com/research/latent-diffusion-models/)] |
+| [RAM++](https://arxiv.org/abs/2310.15200) | ![](https://github.com/xinyu1205/recognize-anything/blob/main/images/ram_plus_compare.jpg) | RAM++ is the next generation of RAM, which can recognize any category with high accuracy. | [[Github](https://github.com/OPPOMKLab/recognize-anything)] |
+| [RAM](https://recognize-anything.github.io/) | ![](https://github.com/xinyu1205/Tag2Text/raw/main/images/localization_and_recognition.jpg) | RAM is an image tagging model, which can recognize any common category with high accuracy. | [[Github](https://github.com/OPPOMKLab/recognize-anything)] <br> [[Demo](https://huggingface.co/spaces/xinyu1205/Recognize_Anything-Tag2Text)] |
+| [BLIP](https://arxiv.org/abs/2201.12086) | ![](https://github.com/salesforce/LAVIS/raw/main/docs/_static/logo_final.png) | A wonderful language-vision model for image understanding. | [[GitHub](https://github.com/salesforce/LAVIS)] |
+| [Visual ChatGPT](https://arxiv.org/abs/2303.04671) | ![](https://github.com/microsoft/TaskMatrix/raw/main/assets/figure.jpg) | A wonderful tool that connects ChatGPT and a series of Visual Foundation Models to enable sending and receiving images during chatting. | [[Github](https://github.com/microsoft/TaskMatrix)] <br> [[Demo](https://huggingface.co/spaces/microsoft/visual_chatgpt)] |
+| [Tag2Text](https://tag2text.github.io/) | ![](https://github.com/xinyu1205/Tag2Text/raw/main/images/tag2text_framework.png) | An efficient and controllable vision-language model which can simultaneously output superior image captioning and image tagging. | [[Github](https://github.com/OPPOMKLab/recognize-anything)] <br> [[Demo](https://huggingface.co/spaces/xinyu1205/Tag2Text)] |
+| [VoxelNeXt](https://arxiv.org/abs/2303.11301) | ![](https://github.com/dvlab-research/VoxelNeXt/raw/master/docs/sequence-v2.gif) | A clean, simple, and fully-sparse 3D object detector, which predicts objects directly upon sparse voxel features. | [[Github](https://github.com/dvlab-research/VoxelNeXt)] 
+
+</div>
+
+## Highlighted Projects
+
+Here we provide some impressive works you may find interesting:
+
+<div align="center">
+
+| Title | Description | Links |
+|:---:|:---:|:---:|
+| [Semantic-SAM](https://github.com/UX-Decoder/Semantic-SAM) | A universal image segmentation model to enable segment and recognize anything at any desired granularity | [[Github](https://github.com/UX-Decoder/Semantic-SAM)] <br> [[Demo](https://github.com/UX-Decoder/Semantic-SAM)] |
+| [SEEM: Segment Everything Everywhere All at Once](https://arxiv.org/pdf/2304.06718.pdf) | A powerful promptable segmentation model supports segmenting with various types of prompts (text, point, scribble, referring image, etc.) and any combination of prompts. | [[Github](https://github.com/UX-Decoder/Segment-Everything-Everywhere-All-At-Once)] <br> [[Demo](https://huggingface.co/spaces/xdecoder/SEEM)] |
+| [OpenSeeD](https://arxiv.org/pdf/2303.08131.pdf) | A simple framework for open-vocabulary segmentation and detection which supports interactive segmentation with box input to generate mask | [[Github](https://github.com/IDEA-Research/OpenSeeD)] |
+| [LLaVA](https://arxiv.org/abs/2304.08485) | Visual instruction tuning with GPT-4 | [[Github](https://github.com/haotian-liu/LLaVA)] <br> [[Page](https://llava-vl.github.io/)] <br> [[Demo](https://llava.hliu.cc/)] <br> [[Data](https://huggingface.co/datasets/liuhaotian/LLaVA-Instruct-150K)] <br> [[Model](https://huggingface.co/liuhaotian/LLaVA-13b-delta-v0)] |
+| [GenSAM](https://arxiv.org/abs/2312.07374) | Relaxing the instance-specific manual prompt requirement in SAM through training-free test-time adaptation | [[Github](https://github.com/jyLin8100/GenSAM)] <br> [[Page](https://lwpyh.github.io/GenSAM/)] |
+
+</div>
+
+We also list some awesome segment-anything extension projects here you may find interesting:
+- [Computer Vision in the Wild (CVinW) Readings](https://github.com/Computer-Vision-in-the-Wild/CVinW_Readings) for those who are interested in open-set tasks in computer vision.
+- [Zero-Shot Anomaly Detection](https://github.com/caoyunkang/GroundedSAM-zero-shot-anomaly-detection) by Yunkang Cao
+- [EditAnything: ControlNet + StableDiffusion based on the SAM segmentation mask](https://github.com/sail-sg/EditAnything) by Shanghua Gao and Pan Zhou
+- [IEA: Image Editing Anything](https://github.com/feizc/IEA) by Zhengcong Fei
+- [SAM-MMRorate: Combining Rotated Object Detector and SAM](https://github.com/Li-Qingyun/sam-mmrotate) by Qingyun Li and Xue Yang
+- [Awesome-Anything](https://github.com/VainF/Awesome-Anything) by Gongfan Fang
+- [Prompt-Segment-Anything](https://github.com/RockeyCoss/Prompt-Segment-Anything) by Rockey
+- [WebUI for Segment-Anything and Grounded-SAM](https://github.com/continue-revolution/sd-webui-segment-anything) by Chengsong Zhang
+- [Inpainting Anything: Inpaint Anything with SAM + Inpainting models](https://github.com/geekyutao/Inpaint-Anything) by Tao Yu
+- [Grounded Segment Anything From Objects to Parts: Combining Segment-Anything with VLPart & GLIP & Visual ChatGPT](https://github.com/Cheems-Seminar/segment-anything-and-name-it) by Peize Sun and Shoufa Chen
+- [Narapi-SAM: Integration of Segment Anything into Narapi (A nice viewer for SAM)](https://github.com/MIC-DKFZ/napari-sam) by MIC-DKFZ
+- [Grounded Segment Anything Colab](https://github.com/camenduru/grounded-segment-anything-colab) by camenduru
+- [Optical Character Recognition with Segment Anything](https://github.com/yeungchenwa/OCR-SAM) by Zhenhua Yang
+- [Transform Image into Unique Paragraph with ChatGPT, BLIP2, OFA, GRIT, Segment Anything, ControlNet](https://github.com/showlab/Image2Paragraph) by showlab
+- [Lang-Segment-Anything: Another awesome demo for combining GroundingDINO with Segment-Anything](https://github.com/luca-medeiros/lang-segment-anything) by Luca Medeiros
+- [🥳 🚀 **Playground: Integrate SAM and OpenMMLab!**](https://github.com/open-mmlab/playground)
+- [3D-object via Segment Anything](https://github.com/dvlab-research/3D-Box-Segment-Anything) by Yukang Chen
+- [Image2Paragraph: Transform Image Into Unique Paragraph](https://github.com/showlab/Image2Paragraph) by Show Lab
+- [Zero-shot Scene Graph Generate with Grounded-SAM](https://github.com/showlab/Image2Paragraph) by JackWhite-rwx
+- [CLIP Surgery for Better Explainability with Enhancement in Open-Vocabulary Tasks](https://github.com/xmed-lab/CLIP_Surgery) by Eli-YiLi
+- [Panoptic-Segment-Anything: Zero-shot panoptic segmentation using SAM](https://github.com/segments-ai/panoptic-segment-anything) by segments-ai
+- [Caption-Anything: Generates Descriptive Captions for Any Object within an Image](https://github.com/ttengwang/Caption-Anything) by Teng Wang
+- [Segment-Anything-3D: Transferring Segmentation Information of 2D Images to 3D Space](https://github.com/Pointcept/SegmentAnything3D) by Yunhan Yang
+- [Expediting SAM without Fine-tuning](https://github.com/Expedit-LargeScale-Vision-Transformer/Expedit-SAM) by Weicong Liang and Yuhui Yuan
+- [Semantic Segment Anything: Providing Rich Semantic Category Annotations for SAM](https://github.com/fudan-zvg/Semantic-Segment-Anything) by Jiaqi Chen and Zeyu Yang and Li Zhang
+- [Enhance Everything: Combining SAM with Image Restoration and Enhancement Tasks](https://github.com/lixinustc/Enhance-Anything) by Xin Li
+- [DragGAN](https://github.com/Zeqiang-Lai/DragGAN) by Shanghai AI Lab.
+- [Tabletop HandyBot: Robotic arm assistant that performs tabletop tasks using Grounded-SAM](https://github.com/ycheng517/tabletop-handybot) by Yifei Cheng
+
+## Installation
+The code requires `python>=3.8`, as well as `pytorch>=1.7` and `torchvision>=0.8`. Please follow the instructions [here](https://pytorch.org/get-started/locally/) to install both PyTorch and TorchVision dependencies. Installing both PyTorch and TorchVision with CUDA support is strongly recommended.
+
+### Install with Docker
+
+Open one terminal:
+
+```
+make build-image
+```
+
+```
+make run
+```
+
+That's it.
+
+If you would like to allow visualization across docker container, open another terminal and type:
+
+```
+xhost +
+```
+
+
+### Install without Docker
+You should set the environment variable manually as follows if you want to build a local GPU environment for Grounded-SAM:
+```bash
+export AM_I_DOCKER=False
+export BUILD_WITH_CUDA=True
+export CUDA_HOME=/path/to/cuda-11.3/
+```
+
+Install Segment Anything:
+
+```bash
+python -m pip install -e segment_anything
+```
+
+Install Grounding DINO:
+
+```bash
+pip install --no-build-isolation -e GroundingDINO
+```
+
+
+Install diffusers:
+
+```bash
+pip install --upgrade diffusers[torch]
+```
+
+Install osx:
+
+```bash
+git submodule update --init --recursive
+cd grounded-sam-osx && bash install.sh
+```
+
+Install RAM & Tag2Text:
+
+```bash
+git clone https://github.com/xinyu1205/recognize-anything.git
+pip install -r ./recognize-anything/requirements.txt
+pip install -e ./recognize-anything/
+```
+
+The following optional dependencies are necessary for mask post-processing, saving masks in COCO format, the example notebooks, and exporting the model in ONNX format. `jupyter` is also required to run the example notebooks.
+
+```
+pip install opencv-python pycocotools matplotlib onnxruntime onnx ipykernel
+```
+
+More details can be found in [install segment anything](https://github.com/facebookresearch/segment-anything#installation) and [install GroundingDINO](https://github.com/IDEA-Research/GroundingDINO#install) and [install OSX](https://github.com/IDEA-Research/OSX)
+
+
+## Grounded-SAM Playground
+Let's start exploring our Grounding-SAM Playground and we will release more interesting demos in the future, stay tuned!
+
+## :open_book: Step-by-Step Notebook Demo
+Here we list some notebook demo provided in this project:
+- [grounded_sam.ipynb](grounded_sam.ipynb)
+- [grounded_sam_colab_demo.ipynb](grounded_sam_colab_demo.ipynb)
+- [grounded_sam_3d_box.ipynb](grounded_sam_3d_box)
+
+
+### :running_man: GroundingDINO: Detect Everything with Text Prompt
+
+:grapes: [[arXiv Paper](https://arxiv.org/abs/2303.05499)] &nbsp; :rose:[[Try the Colab Demo](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/zero-shot-object-detection-with-grounding-dino.ipynb)] &nbsp; :sunflower: [[Try Huggingface Demo](https://huggingface.co/spaces/ShilongLiu/Grounding_DINO_demo)] &nbsp; :mushroom: [[Automated Dataset Annotation and Evaluation](https://youtu.be/C4NqaRBz_Kw)]
+
+Here's the step-by-step tutorial on running `GroundingDINO` demo:
+
+**Step 1: Download the pretrained weights**
+
+```bash
+cd Grounded-Segment-Anything
+
+# download the pretrained groundingdino-swin-tiny model
+wget https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth
+```
+
+**Step 2: Running the demo**
+
+```bash
+python grounding_dino_demo.py
+```
+
+<details>
+<summary> <b> Running with Python (same as demo but you can run it anywhere after installing GroundingDINO) </b> </summary>
+
+```python
+from groundingdino.util.inference import load_model, load_image, predict, annotate
+import cv2
+
+model = load_model("GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py", "./groundingdino_swint_ogc.pth")
+IMAGE_PATH = "assets/demo1.jpg"
+TEXT_PROMPT = "bear."
+BOX_THRESHOLD = 0.35
+TEXT_THRESHOLD = 0.25
+
+image_source, image = load_image(IMAGE_PATH)
+
+boxes, logits, phrases = predict(
+    model=model,
+    image=image,
+    caption=TEXT_PROMPT,
+    box_threshold=BOX_THRESHOLD,
+    text_threshold=TEXT_THRESHOLD
+)
+
+annotated_frame = annotate(image_source=image_source, boxes=boxes, logits=logits, phrases=phrases)
+cv2.imwrite("annotated_image.jpg", annotated_frame)
+```
+
+</details>
+<br>
+
+**Tips**
+- If you want to detect multiple objects in one sentence with [Grounding DINO](https://github.com/IDEA-Research/GroundingDINO), we suggest separating each name with `.` . An example: `cat . dog . chair .`
+
+**Step 3: Check the annotated image**
+
+The annotated image will be saved as `./annotated_image.jpg`.
+
+<div align="center">
+
+| Text Prompt | Demo Image | Annotated Image |
+|:----:|:----:|:----:|
+| `Bear.` | ![](./assets/demo1.jpg)  | ![](./assets/annotated_image.jpg) |
+| `Horse. Clouds. Grasses. Sky. Hill` | ![](./assets/demo7.jpg)  | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounding_dino/groundingdino_demo7.jpg?raw=true)
+
+</div>
+
+
+### :running_man: Grounded-SAM: Detect and Segment Everything with Text Prompt
+
+Here's the step-by-step tutorial on running `Grounded-SAM` demo:
+
+**Step 1: Download the pretrained weights**
+
+```bash
+cd Grounded-Segment-Anything
+
+wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth
+wget https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth
+```
+
+We provide two versions of Grounded-SAM demo here:
+- [grounded_sam_demo.py](./grounded_sam_demo.py): our original implementation for Grounded-SAM.
+- [grounded_sam_simple_demo.py](./grounded_sam_simple_demo.py) our updated more elegant version for Grounded-SAM.
+
+**Step 2: Running original grounded-sam demo**
+```bash
+# depends on your device 
+export CUDA_VISIBLE_DEVICES=0
+```
+
+```python
+
+python grounded_sam_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/demo1.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --text_prompt "bear" \
+  --device "cuda"
+```
+
+The annotated results will be saved in `./outputs` as follows
+
+<div align="center">
+
+| Input Image | Annotated Image | Generated Mask |
+|:----:|:----:|:----:|
+| ![](./assets/demo1.jpg) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounded_sam/original_grounded_sam_demo1.jpg?raw=true) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounded_sam/mask.jpg?raw=true) |
+
+</div>
+
+**Step 3: Running grounded-sam demo with sam-hq**
+- Download the demo image
+```bash
+wget https://github.com/IDEA-Research/detrex-storage/releases/download/grounded-sam-storage/sam_hq_demo_image.png
+```
+
+- Download SAM-HQ checkpoint [here](https://github.com/SysCV/sam-hq#model-checkpoints)
+
+- Running grounded-sam-hq demo as follows:
+```python
+export CUDA_VISIBLE_DEVICES=0
+python grounded_sam_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_hq_checkpoint ./sam_hq_vit_h.pth \  # path to sam-hq checkpoint
+  --use_sam_hq \  # set to use sam-hq model
+  --input_image sam_hq_demo_image.png \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --text_prompt "chair." \
+  --device "cuda"
+```
+
+The annotated results will be saved in `./outputs` as follows
+
+<div align="center">
+
+| Input Image | SAM Output | SAM-HQ Output |
+|:----:|:----:|:----:|
+| ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/sam_hq/sam_hq_demo.png?raw=true) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/sam_hq/sam_output.jpg?raw=true) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/sam_hq/sam_hq_output.jpg?raw=true) |
+
+</div>
+
+**Step 4: Running the updated grounded-sam demo (optional)**
+
+Note that this demo is almost same as the original demo, but **with more elegant code**.
+
+```python
+python grounded_sam_simple_demo.py
+```
+
+The annotated results will be saved as `./groundingdino_annotated_image.jpg` and `./grounded_sam_annotated_image.jpg`
+
+<div align="center">
+
+| Text Prompt | Input Image | GroundingDINO Annotated Image | Grounded-SAM Annotated Image |
+|:----:|:----:|:----:|:----:|
+| `The running dog` | ![](./assets/demo2.jpg) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounded_sam/groundingdino_annotated_image_demo2.jpg?raw=true) | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounded_sam/grounded_sam_annotated_image_demo2.jpg?raw=true) |
+| `Horse. Clouds. Grasses. Sky. Hill` | ![](./assets/demo7.jpg) | ![](assets/groundingdino_annotated_image.jpg) | ![](assets/grounded_sam_annotated_image.jpg) |
+
+</div>
+
+**Step 5: Running the Sam model with multi-gpu**
+```bash
+export CUDA_VISIBLE_DEVICES=0,1
+```
+```python
+
+python grounded_sam_multi_gpu_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_path assets/car \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --text_prompt "car" \
+  --device "cuda"
+```
+You will see that the model is loaded once per GPU ![](assets/multi-gpu.png) 
+
+### :skier: Grounded-SAM with Inpainting: Detect, Segment and Generate Everything with Text Prompt
+
+**Step 1: Download the pretrained weights**
+
+```bash
+cd Grounded-Segment-Anything
+
+wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth
+wget https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth
+```
+
+**Step 2: Running grounded-sam inpainting demo**
+
+```bash
+CUDA_VISIBLE_DEVICES=0
+python grounded_sam_inpainting_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/inpaint_demo.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --det_prompt "bench" \
+  --inpaint_prompt "A sofa, high quality, detailed" \
+  --device "cuda"
+```
+
+The annotated and inpaint image will be saved in `./outputs`
+
+**Step 3: Check the results**
+
+
+<div align="center">
+
+| Input Image | Det Prompt | Annotated Image | Inpaint Prompt | Inpaint Image |
+|:---:|:---:|:---:|:---:|:---:|
+|![](./assets/inpaint_demo.jpg) | `Bench` | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounded_sam_inpaint/grounded_sam_output.jpg?raw=true) | `A sofa, high quality, detailed` | ![](https://github.com/IDEA-Research/detrex-storage/blob/main/assets/grounded_sam/grounded_sam_inpaint/grounded_sam_inpainting_output.jpg?raw=true) |
+
+</div>
+
+### :golfing: Grounded-SAM and Inpaint Gradio APP
+
+We support 6 tasks in the local Gradio APP：
+
+1. **scribble**: Segmentation is achieved through Segment Anything and mouse click interaction (you need to click on the object with the mouse, no need to specify the prompt).
+2. **automask**: Segment the entire image at once through Segment Anything (no need to specify a prompt).
+3. **det**: Realize detection through Grounding DINO and text interaction (text prompt needs to be specified).
+4. **seg**: Realize text interaction by combining Grounding DINO and Segment Anything to realize detection + segmentation (need to specify text prompt).
+5. **inpainting**: By combining Grounding DINO + Segment Anything + Stable Diffusion to achieve text exchange and replace the target object (need to specify text prompt and inpaint prompt) .
+6. **automatic**: By combining BLIP + Grounding DINO + Segment Anything to achieve non-interactive detection + segmentation (no need to specify prompt).
+
+```bash
+python gradio_app.py
+```
+
+- The gradio_app visualization as follows:
+
+![](./assets/gradio_demo.png)
+
+
+### :label: Grounded-SAM with RAM or Tag2Text for Automatic Labeling
+[**The Recognize Anything Models**](https://github.com/OPPOMKLab/recognize-anything) are a series of open-source and strong fundamental image recognition models, including [RAM++](https://arxiv.org/abs/2310.15200), [RAM](https://arxiv.org/abs/2306.03514) and [Tag2text](https://arxiv.org/abs/2303.05657).
+
+
+It is seamlessly linked to generate pseudo labels automatically as follows:
+1. Use RAM/Tag2Text to generate tags.
+2. Use Grounded-Segment-Anything to generate the boxes and masks.
+
+
+**Step 1: Init submodule and download the pretrained checkpoint**
+
+- Init submodule:
+
+```bash
+cd Grounded-Segment-Anything
+git submodule init
+git submodule update
+```
+
+- Download pretrained weights for `GroundingDINO`, `SAM` and `RAM/Tag2Text`:
+
+```bash
+wget https://dl.fbaipublicfiles.com/segment_anything/sam_vit_h_4b8939.pth
+wget https://github.com/IDEA-Research/GroundingDINO/releases/download/v0.1.0-alpha/groundingdino_swint_ogc.pth
+
+
+wget https://huggingface.co/spaces/xinyu1205/Tag2Text/resolve/main/ram_swin_large_14m.pth
+wget https://huggingface.co/spaces/xinyu1205/Tag2Text/resolve/main/tag2text_swin_14m.pth
+```
+
+**Step 2: Running the demo with RAM**
+```bash
+export CUDA_VISIBLE_DEVICES=0
+python automatic_label_ram_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --ram_checkpoint ram_swin_large_14m.pth \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/demo9.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.25 \
+  --text_threshold 0.2 \
+  --iou_threshold 0.5 \
+  --device "cuda"
+```
+
+
+**Step 2: Or Running the demo with Tag2Text**
+```bash
+export CUDA_VISIBLE_DEVICES=0
+python automatic_label_tag2text_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --tag2text_checkpoint tag2text_swin_14m.pth \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/demo9.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.25 \
+  --text_threshold 0.2 \
+  --iou_threshold 0.5 \
+  --device "cuda"
+```
+
+- RAM++ significantly improves the open-set capability of RAM, for [RAM++ inference on unseen categoreis](https://github.com/xinyu1205/recognize-anything#ram-inference-on-unseen-categories-open-set).
+- Tag2Text also provides powerful captioning capabilities, and the process with captions can refer to [BLIP](#robot-run-grounded-segment-anything--blip-demo).
+- The pseudo labels and model prediction visualization will be saved in `output_dir` as follows (right figure):
+
+![](./assets/automatic_label_output/demo9_tag2text_ram.jpg)
+
+
+### :robot: Grounded-SAM with BLIP for Automatic Labeling
+It is easy to generate pseudo labels automatically as follows:
+1. Use BLIP (or other caption models) to generate a caption.
+2. Extract tags from the caption. We use ChatGPT to handle the potential complicated sentences. 
+3. Use Grounded-Segment-Anything to generate the boxes and masks.
+
+- Run Demo
+```bash
+export OPENAI_API_KEY=your_openai_key
+export OPENAI_API_BASE=https://closeai.deno.dev/v1
+export CUDA_VISIBLE_DEVICES=0
+python automatic_label_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/demo3.jpg \
+  --output_dir "outputs" \
+  --openai_key $OPENAI_API_KEY \
+  --box_threshold 0.25 \
+  --text_threshold 0.2 \
+  --iou_threshold 0.5 \
+  --device "cuda"
+```
+
+- When you don't have a paid Account for ChatGPT is also possible to use NLTK instead. Just don't include the ```openai_key``` Parameter when starting the Demo.
+  - The Script will automatically download the necessary NLTK Data.
+- The pseudo labels and model prediction visualization will be saved in `output_dir` as follows:
+
+![](./assets/automatic_label_output_demo3.jpg)
+
+
+### :open_mouth: Grounded-SAM with Whisper: Detect and Segment Anything with Audio
+Detect and segment anything with speech!
+
+![](assets/acoustics/gsam_whisper_inpainting_demo.png)
+
+**Install Whisper**
+```bash
+pip install -U openai-whisper
+```
+See the [whisper official page](https://github.com/openai/whisper#setup) if you have other questions for the installation.
+
+**Run Voice-to-Label Demo**
+
+Optional: Download the demo audio file
+
+```bash
+wget https://huggingface.co/ShilongLiu/GroundingDINO/resolve/main/demo_audio.mp3
+```
+
+
+```bash
+export CUDA_VISIBLE_DEVICES=0
+python grounded_sam_whisper_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/demo4.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --speech_file "demo_audio.mp3" \
+  --device "cuda"
+```
+
+![](./assets/grounded_sam_whisper_output.jpg)
+
+**Run Voice-to-inpaint Demo**
+
+You can enable chatgpt to help you automatically detect the object and inpainting order with `--enable_chatgpt`. 
+
+Or you can specify the object you want to inpaint [stored in `args.det_speech_file`] and the text you want to inpaint with [stored in `args.inpaint_speech_file`].
+
+```bash
+export OPENAI_API_KEY=your_openai_key
+export OPENAI_API_BASE=https://closeai.deno.dev/v1
+# Example: enable chatgpt
+export CUDA_VISIBLE_DEVICES=0
+python grounded_sam_whisper_inpainting_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/inpaint_demo.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --prompt_speech_file assets/acoustics/prompt_speech_file.mp3 \
+  --enable_chatgpt \
+  --openai_key $OPENAI_API_KEY\
+  --device "cuda"
+```
+
+```bash
+# Example: without chatgpt
+export CUDA_VISIBLE_DEVICES=0
+python grounded_sam_whisper_inpainting_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --input_image assets/inpaint_demo.jpg \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --det_speech_file "assets/acoustics/det_voice.mp3" \
+  --inpaint_speech_file "assets/acoustics/inpaint_voice.mp3" \
+  --device "cuda"
+```
+
+![](./assets/acoustics/gsam_whisper_inpainting_pipeline.png)
+
+### :speech_balloon: Grounded-SAM ChatBot Demo
+
+https://user-images.githubusercontent.com/24236723/231955561-2ae4ec1a-c75f-4cc5-9b7b-517aa1432123.mp4
+
+Following [Visual ChatGPT](https://github.com/microsoft/visual-chatgpt), we add a ChatBot for our project. Currently, it supports:
+1. "Describe the image."
+2. "Detect the dog (and the cat) in the image."
+3. "Segment anything in the image."
+4. "Segment the dog (and the cat) in the image."
+5. "Help me label the image."
+6. "Replace the dog with a cat in the image."
+
+To use the ChatBot:
+- Install whisper if you want to use audio as input.
+- Set the default model setting in the tool `Grounded_dino_sam_inpainting`.
+- Run Demo
+```bash
+export OPENAI_API_KEY=your_openai_key
+export OPENAI_API_BASE=https://closeai.deno.dev/v1
+export CUDA_VISIBLE_DEVICES=0
+python chatbot.py 
+```
+
+### :man_dancing: Run Grounded-Segment-Anything + OSX Demo
+
+<p align="middle">
+<img src="assets/osx/grouned_sam_osx_demo.gif">
+<br>
+</p>
+
+
+- Download the checkpoint `osx_l_wo_decoder.pth.tar` from [here](https://drive.google.com/drive/folders/1x7MZbB6eAlrq5PKC9MaeIm4GqkBpokow?usp=share_link) for OSX:
+- Download the human model files and place it into `grounded-sam-osx/utils/human_model_files` following the instruction of [OSX](https://github.com/IDEA-Research/OSX).
+
+- Run Demo
+
+```shell
+export CUDA_VISIBLE_DEVICES=0
+python grounded_sam_osx_demo.py \
+  --config GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py \
+  --grounded_checkpoint groundingdino_swint_ogc.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --osx_checkpoint osx_l_wo_decoder.pth.tar \
+  --input_image assets/osx/grounded_sam_osx_demo.png \
+  --output_dir "outputs" \
+  --box_threshold 0.3 \
+  --text_threshold 0.25 \
+  --text_prompt "humans, chairs" \
+  --device "cuda"
+```
+
+- The model prediction visualization will be saved in `output_dir` as follows:
+
+<img src="assets/osx/grounded_sam_osx_output.jpg" style="zoom: 49%;" />
+
+- We also support promptable 3D whole-body mesh recovery. For example, you can track someone with a text prompt  and estimate his 3D pose and shape :
+
+| ![space-1.jpg](assets/osx/grounded_sam_osx_output1.jpg) |
+| :---------------------------------------------------: |
+|             *A person with pink clothes*              |
+
+| ![space-1.jpg](assets/osx/grounded_sam_osx_output2.jpg) |
+| :---------------------------------------------------: |
+|               *A man with a sunglasses*               |
+
+
+## :man_dancing: Run Grounded-Segment-Anything + VISAM Demo
+
+- Download the checkpoint `motrv2_dancetrack.pth` from [here](https://drive.google.com/file/d/1EA4lndu2yQcVgBKR09KfMe5efbf631Th/view?usp=share_link) for MOTRv2:
+- See the more thing if you have other questions for the installation.
+
+- Run Demo
+
+```shell
+export CUDA_VISIBLE_DEVICES=0
+python grounded_sam_visam.py \
+  --meta_arch motr \
+  --dataset_file e2e_dance \
+  --with_box_refine \
+  --query_interaction_layer QIMv2 \
+  --num_queries 10 \
+  --det_db det_db_motrv2.json \
+  --use_checkpoint \
+  --mot_path your_data_path \
+  --resume motrv2_dancetrack.pth \
+  --sam_checkpoint sam_vit_h_4b8939.pth \
+  --video_path DanceTrack/test/dancetrack0003 
+```
+|![](https://raw.githubusercontent.com/BingfengYan/MOTSAM/main/visam.gif)|
+
+
+### :dancers: Interactive Editing
+- Release the interactive fashion-edit playground in [here](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/humanFace). Run in the notebook, just click for annotating points for further segmentation. Enjoy it! 
+
+
+- Release human-face-edit branch [here](https://github.com/IDEA-Research/Grounded-Segment-Anything/tree/humanFace). We'll keep updating this branch with more interesting features. Here are some examples:
+
+  ![](https://github.com/IDEA-Research/Grounded-Segment-Anything/blob/humanFace/assets/231-hair-edit.png)
+
+## :camera: 3D-Box via Segment Anything
+We extend the scope to 3D world by combining Segment Anything and [VoxelNeXt](https://github.com/dvlab-research/VoxelNeXt). When we provide a prompt (e.g., a point / box), the result is not only 2D segmentation mask, but also 3D boxes. Please check [voxelnext_3d_box](./voxelnext_3d_box/) for more details.
+  ![](https://github.com/IDEA-Research/Grounded-Segment-Anything/blob/main/voxelnext_3d_box/images/sam-voxelnext.png)
+  ![](https://github.com/IDEA-Research/Grounded-Segment-Anything/blob/main/voxelnext_3d_box/images/image_boxes2.png)
+
+
+
+
+## :cupid: Acknowledgements
+
+- [Segment Anything](https://github.com/facebookresearch/segment-anything)
+- [Grounding DINO](https://github.com/IDEA-Research/GroundingDINO)
+
+
+## Contributors
+
+Our project wouldn't be possible without the contributions of these amazing people! Thank you all for making this project better.
+
+<a href="https://github.com/IDEA-Research/Grounded-Segment-Anything/graphs/contributors">
+  <img src="https://contrib.rocks/image?repo=IDEA-Research/Grounded-Segment-Anything" />
+</a>
+
+
+## Citation
+If you find this project helpful for your research, please consider citing the following BibTeX entry.
+```BibTex
+@article{kirillov2023segany,
+  title={Segment Anything}, 
+  author={Kirillov, Alexander and Mintun, Eric and Ravi, Nikhila and Mao, Hanzi and Rolland, Chloe and Gustafson, Laura and Xiao, Tete and Whitehead, Spencer and Berg, Alexander C. and Lo, Wan-Yen and Doll{\'a}r, Piotr and Girshick, Ross},
+  journal={arXiv:2304.02643},
+  year={2023}
+}
+
+@article{liu2023grounding,
+  title={Grounding dino: Marrying dino with grounded pre-training for open-set object detection},
+  author={Liu, Shilong and Zeng, Zhaoyang and Ren, Tianhe and Li, Feng and Zhang, Hao and Yang, Jie and Li, Chunyuan and Yang, Jianwei and Su, Hang and Zhu, Jun and others},
+  journal={arXiv preprint arXiv:2303.05499},
+  year={2023}
+}
+
+@misc{ren2024grounded,
+      title={Grounded SAM: Assembling Open-World Models for Diverse Visual Tasks}, 
+      author={Tianhe Ren and Shilong Liu and Ailing Zeng and Jing Lin and Kunchang Li and He Cao and Jiayu Chen and Xinyu Huang and Yukang Chen and Feng Yan and Zhaoyang Zeng and Hao Zhang and Feng Li and Jie Yang and Hongyang Li and Qing Jiang and Lei Zhang},
+      year={2024},
+      eprint={2401.14159},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV}
+}
+```

external/Grounded-Segment-Anything/automatic_label_simple_demo.py

@@ -0,0 +1,166 @@
+import cv2
+import numpy as np
+import supervision as sv
+from typing import List
+from PIL import Image
+
+import torch
+
+from groundingdino.util.inference import Model
+from segment_anything import sam_model_registry, SamPredictor
+
+# Tag2Text
+# from ram.models import tag2text_caption
+from ram.models import ram
+# from ram import inference_tag2text
+from ram import inference_ram
+import torchvision
+import torchvision.transforms as TS
+
+
+# Hyper-Params
+SOURCE_IMAGE_PATH = "./assets/demo9.jpg"
+DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+
+GROUNDING_DINO_CONFIG_PATH = "GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py"
+GROUNDING_DINO_CHECKPOINT_PATH = "./groundingdino_swint_ogc.pth"
+
+SAM_ENCODER_VERSION = "vit_h"
+SAM_CHECKPOINT_PATH = "./sam_vit_h_4b8939.pth"
+
+TAG2TEXT_CHECKPOINT_PATH = "./tag2text_swin_14m.pth"
+RAM_CHECKPOINT_PATH = "./ram_swin_large_14m.pth"
+
+TAG2TEXT_THRESHOLD = 0.64
+BOX_THRESHOLD = 0.2
+TEXT_THRESHOLD = 0.2
+IOU_THRESHOLD = 0.5
+
+# Building GroundingDINO inference model
+grounding_dino_model = Model(model_config_path=GROUNDING_DINO_CONFIG_PATH, model_checkpoint_path=GROUNDING_DINO_CHECKPOINT_PATH)
+
+
+# Building SAM Model and SAM Predictor
+sam = sam_model_registry[SAM_ENCODER_VERSION](checkpoint=SAM_CHECKPOINT_PATH)
+sam_predictor = SamPredictor(sam)
+
+# Tag2Text
+# initialize Tag2Text
+normalize = TS.Normalize(
+    mean=[0.485, 0.456, 0.406],
+    std=[0.229, 0.224, 0.225]
+)
+transform = TS.Compose(
+    [
+        TS.Resize((384, 384)),
+        TS.ToTensor(), 
+        normalize
+    ]
+)
+
+DELETE_TAG_INDEX = []  # filter out attributes and action which are difficult to be grounded
+for idx in range(3012, 3429):
+    DELETE_TAG_INDEX.append(idx)
+
+# tag2text_model = tag2text_caption(
+#     pretrained=TAG2TEXT_CHECKPOINT_PATH,
+#     image_size=384,
+#     vit='swin_b',
+#     delete_tag_index=DELETE_TAG_INDEX
+# )
+# # threshold for tagging
+# # we reduce the threshold to obtain more tags
+# tag2text_model.threshold = TAG2TEXT_THRESHOLD
+# tag2text_model.eval()
+# tag2text_model = tag2text_model.to(DEVICE)
+
+ram_model = ram(pretrained=RAM_CHECKPOINT_PATH,
+                                        image_size=384,
+                                        vit='swin_l')
+ram_model.eval()
+ram_model = ram_model.to(DEVICE)
+
+# load image
+image = cv2.imread(SOURCE_IMAGE_PATH)  # bgr
+image_pillow = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))  # rgb
+
+image_pillow = image_pillow.resize((384, 384))
+image_pillow = transform(image_pillow).unsqueeze(0).to(DEVICE)
+
+specified_tags='None'
+# res = inference_tag2text(image_pillow , tag2text_model, specified_tags)
+res = inference_ram(image_pillow , ram_model)
+
+# Currently ", " is better for detecting single tags
+# while ". " is a little worse in some case
+AUTOMATIC_CLASSES=res[0].split(" | ")
+
+print(f"Tags: {res[0].replace(' |', ',')}")
+
+
+# detect objects
+detections = grounding_dino_model.predict_with_classes(
+    image=image,
+    classes=AUTOMATIC_CLASSES,
+    box_threshold=BOX_THRESHOLD,
+    text_threshold=BOX_THRESHOLD
+)
+
+# NMS post process
+print(f"Before NMS: {len(detections.xyxy)} boxes")
+nms_idx = torchvision.ops.nms(
+    torch.from_numpy(detections.xyxy), 
+    torch.from_numpy(detections.confidence), 
+    IOU_THRESHOLD
+).numpy().tolist()
+
+detections.xyxy = detections.xyxy[nms_idx]
+detections.confidence = detections.confidence[nms_idx]
+detections.class_id = detections.class_id[nms_idx]
+
+print(f"After NMS: {len(detections.xyxy)} boxes")
+
+# annotate image with detections
+box_annotator = sv.BoxAnnotator()
+labels = [
+    f"{AUTOMATIC_CLASSES[class_id]} {confidence:0.2f}" 
+    for _, _, confidence, class_id, _, _
+    in detections]
+annotated_frame = box_annotator.annotate(scene=image.copy(), detections=detections, labels=labels)
+
+# save the annotated grounding dino image
+cv2.imwrite("groundingdino_auto_annotated_image.jpg", annotated_frame)
+
+# Prompting SAM with detected boxes
+def segment(sam_predictor: SamPredictor, image: np.ndarray, xyxy: np.ndarray) -> np.ndarray:
+    sam_predictor.set_image(image)
+    result_masks = []
+    for box in xyxy:
+        masks, scores, logits = sam_predictor.predict(
+            box=box,
+            multimask_output=True
+        )
+        index = np.argmax(scores)
+        result_masks.append(masks[index])
+    return np.array(result_masks)
+
+
+# convert detections to masks
+detections.mask = segment(
+    sam_predictor=sam_predictor,
+    image=cv2.cvtColor(image, cv2.COLOR_BGR2RGB),
+    xyxy=detections.xyxy
+)
+
+# annotate image with detections
+box_annotator = sv.BoxAnnotator()
+mask_annotator = sv.MaskAnnotator()
+labels = [
+    f"{AUTOMATIC_CLASSES[class_id]} {confidence:0.2f}" 
+    for _, _, confidence, class_id, _, _
+    in detections]
+annotated_image = mask_annotator.annotate(scene=image.copy(), detections=detections)
+annotated_image = box_annotator.annotate(scene=annotated_image, detections=detections, labels=labels)
+
+# save the annotated grounded-sam image
+cv2.imwrite("ram_grounded_sam_auto_annotated_image.jpg", annotated_image)

external/Grounded-Segment-Anything/grounded_sam_demo.py

@@ -0,0 +1,242 @@
+import argparse
+import os
+import sys
+
+import numpy as np
+import json
+import torch
+from PIL import Image
+
+sys.path.append(os.path.join(os.getcwd(), "GroundingDINO"))
+sys.path.append(os.path.join(os.getcwd(), "segment_anything"))
+
+
+# Grounding DINO
+import GroundingDINO.groundingdino.datasets.transforms as T
+from GroundingDINO.groundingdino.models import build_model
+from GroundingDINO.groundingdino.util.slconfig import SLConfig
+from GroundingDINO.groundingdino.util.utils import clean_state_dict, get_phrases_from_posmap
+
+
+# segment anything
+from segment_anything import (
+    sam_model_registry,
+    sam_hq_model_registry,
+    SamPredictor
+)
+import cv2
+import numpy as np
+import matplotlib.pyplot as plt
+
+
+def load_image(image_path):
+    # load image
+    image_pil = Image.open(image_path).convert("RGB")  # load image
+
+    transform = T.Compose(
+        [
+            T.RandomResize([800], max_size=1333),
+            T.ToTensor(),
+            T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
+        ]
+    )
+    image, _ = transform(image_pil, None)  # 3, h, w
+    return image_pil, image
+
+
+def load_model(model_config_path, model_checkpoint_path, bert_base_uncased_path, device):
+    args = SLConfig.fromfile(model_config_path)
+    args.device = device
+    args.bert_base_uncased_path = bert_base_uncased_path
+    model = build_model(args)
+    checkpoint = torch.load(model_checkpoint_path, map_location="cpu")
+    load_res = model.load_state_dict(clean_state_dict(checkpoint["model"]), strict=False)
+    print(load_res)
+    _ = model.eval()
+    return model
+
+
+def get_grounding_output(model, image, caption, box_threshold, text_threshold, with_logits=True, device="cpu"):
+    caption = caption.lower()
+    caption = caption.strip()
+    if not caption.endswith("."):
+        caption = caption + "."
+    model = model.to(device)
+    image = image.to(device)
+    with torch.no_grad():
+        outputs = model(image[None], captions=[caption])
+    logits = outputs["pred_logits"].cpu().sigmoid()[0]  # (nq, 256)
+    boxes = outputs["pred_boxes"].cpu()[0]  # (nq, 4)
+    logits.shape[0]
+
+    # filter output
+    logits_filt = logits.clone()
+    boxes_filt = boxes.clone()
+    filt_mask = logits_filt.max(dim=1)[0] > box_threshold
+    logits_filt = logits_filt[filt_mask]  # num_filt, 256
+    boxes_filt = boxes_filt[filt_mask]  # num_filt, 4
+    logits_filt.shape[0]
+
+    # get phrase
+    tokenlizer = model.tokenizer
+    tokenized = tokenlizer(caption)
+    # build pred
+    pred_phrases = []
+    for logit, box in zip(logits_filt, boxes_filt):
+        pred_phrase = get_phrases_from_posmap(logit > text_threshold, tokenized, tokenlizer)
+        if with_logits:
+            pred_phrases.append(pred_phrase + f"({str(logit.max().item())[:4]})")
+        else:
+            pred_phrases.append(pred_phrase)
+
+    return boxes_filt, pred_phrases
+
+def show_mask(mask, ax, random_color=False):
+    if random_color:
+        color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
+    else:
+        color = np.array([30/255, 144/255, 255/255, 0.6])
+    h, w = mask.shape[-2:]
+    mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
+    ax.imshow(mask_image)
+
+
+def show_box(box, ax, label):
+    x0, y0 = box[0], box[1]
+    w, h = box[2] - box[0], box[3] - box[1]
+    ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor='green', facecolor=(0,0,0,0), lw=2))
+    ax.text(x0, y0, label)
+
+
+def save_mask_data(output_dir, mask_list, box_list, label_list):
+    value = 0  # 0 for background
+
+    mask_img = torch.zeros(mask_list.shape[-2:])
+    for idx, mask in enumerate(mask_list):
+        mask_img[mask.cpu().numpy()[0] == True] = value + idx + 1
+    plt.figure(figsize=(10, 10))
+    plt.imshow(mask_img.numpy())
+    plt.axis('off')
+    plt.savefig(os.path.join(output_dir, 'mask.jpg'), bbox_inches="tight", dpi=300, pad_inches=0.0)
+
+    json_data = [{
+        'value': value,
+        'label': 'background'
+    }]
+    for label, box in zip(label_list, box_list):
+        value += 1
+        name, logit = label.split('(')
+        logit = logit[:-1] # the last is ')'
+        json_data.append({
+            'value': value,
+            'label': name,
+            'logit': float(logit),
+            'box': box.numpy().tolist(),
+        })
+    with open(os.path.join(output_dir, 'mask.json'), 'w') as f:
+        json.dump(json_data, f)
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser("Grounded-Segment-Anything Demo", add_help=True)
+    parser.add_argument("--config", type=str, required=True, help="path to config file")
+    parser.add_argument(
+        "--grounded_checkpoint", type=str, required=True, help="path to checkpoint file"
+    )
+    parser.add_argument(
+        "--sam_version", type=str, default="vit_h", required=False, help="SAM ViT version: vit_b / vit_l / vit_h"
+    )
+    parser.add_argument(
+        "--sam_checkpoint", type=str, required=False, help="path to sam checkpoint file"
+    )
+    parser.add_argument(
+        "--sam_hq_checkpoint", type=str, default=None, help="path to sam-hq checkpoint file"
+    )
+    parser.add_argument(
+        "--use_sam_hq", action="store_true", help="using sam-hq for prediction"
+    )
+    parser.add_argument("--input_image", type=str, required=True, help="path to image file")
+    parser.add_argument("--text_prompt", type=str, required=True, help="text prompt")
+    parser.add_argument(
+        "--output_dir", "-o", type=str, default="outputs", required=True, help="output directory"
+    )
+
+    parser.add_argument("--box_threshold", type=float, default=0.3, help="box threshold")
+    parser.add_argument("--text_threshold", type=float, default=0.25, help="text threshold")
+
+    parser.add_argument("--device", type=str, default="cpu", help="running on cpu only!, default=False")
+    parser.add_argument("--bert_base_uncased_path", type=str, required=False, help="bert_base_uncased model path, default=False")
+    args = parser.parse_args()
+
+    # cfg
+    config_file = args.config  # change the path of the model config file
+    grounded_checkpoint = args.grounded_checkpoint  # change the path of the model
+    sam_version = args.sam_version
+    sam_checkpoint = args.sam_checkpoint
+    sam_hq_checkpoint = args.sam_hq_checkpoint
+    use_sam_hq = args.use_sam_hq
+    image_path = args.input_image
+    text_prompt = args.text_prompt
+    output_dir = args.output_dir
+    box_threshold = args.box_threshold
+    text_threshold = args.text_threshold
+    device = args.device
+    bert_base_uncased_path = args.bert_base_uncased_path
+
+    # make dir
+    os.makedirs(output_dir, exist_ok=True)
+    # load image
+    image_pil, image = load_image(image_path)
+    # load model
+    model = load_model(config_file, grounded_checkpoint, bert_base_uncased_path, device=device)
+
+    # visualize raw image
+    image_pil.save(os.path.join(output_dir, "raw_image.jpg"))
+
+    # run grounding dino model
+    boxes_filt, pred_phrases = get_grounding_output(
+        model, image, text_prompt, box_threshold, text_threshold, device=device
+    )
+
+    # initialize SAM
+    if use_sam_hq:
+        predictor = SamPredictor(sam_hq_model_registry[sam_version](checkpoint=sam_hq_checkpoint).to(device))
+    else:
+        predictor = SamPredictor(sam_model_registry[sam_version](checkpoint=sam_checkpoint).to(device))
+    image = cv2.imread(image_path)
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+    predictor.set_image(image)
+
+    size = image_pil.size
+    H, W = size[1], size[0]
+    for i in range(boxes_filt.size(0)):
+        boxes_filt[i] = boxes_filt[i] * torch.Tensor([W, H, W, H])
+        boxes_filt[i][:2] -= boxes_filt[i][2:] / 2
+        boxes_filt[i][2:] += boxes_filt[i][:2]
+
+    boxes_filt = boxes_filt.cpu()
+    transformed_boxes = predictor.transform.apply_boxes_torch(boxes_filt, image.shape[:2]).to(device)
+
+    masks, _, _ = predictor.predict_torch(
+        point_coords = None,
+        point_labels = None,
+        boxes = transformed_boxes.to(device),
+        multimask_output = False,
+    )
+
+    # draw output image
+    plt.figure(figsize=(10, 10))
+    plt.imshow(image)
+    for mask in masks:
+        show_mask(mask.cpu().numpy(), plt.gca(), random_color=True)
+    for box, label in zip(boxes_filt, pred_phrases):
+        show_box(box.numpy(), plt.gca(), label)
+
+    plt.axis('off')
+    plt.savefig(
+        os.path.join(output_dir, "grounded_sam_output.jpg"),
+        bbox_inches="tight", dpi=300, pad_inches=0.0
+    )
+
+    save_mask_data(output_dir, masks, boxes_filt, pred_phrases)

external/Grounded-Segment-Anything/grounded_sam_osx_demo.py

@@ -0,0 +1,299 @@
+import torchvision.transforms as transforms
+from torch.nn.parallel.data_parallel import DataParallel
+import torch.backends.cudnn as cudnn
+import argparse
+import json
+import torch
+from PIL import Image
+import matplotlib.pyplot as plt
+import os
+import cv2
+import numpy as np
+
+# Grounding DINO
+import GroundingDINO.groundingdino.datasets.transforms as T
+from GroundingDINO.groundingdino.models import build_model
+from GroundingDINO.groundingdino.util.slconfig import SLConfig
+from GroundingDINO.groundingdino.util.utils import clean_state_dict, get_phrases_from_posmap
+
+# segment anything
+from segment_anything import build_sam, SamPredictor
+
+
+# OSX
+import sys
+sys.path.insert(0, 'grounded-sam-osx')
+from osx import get_model
+from config import cfg
+from utils.preprocessing import load_img, process_bbox, generate_patch_image
+from utils.human_models import smpl_x
+
+os.environ["PYOPENGL_PLATFORM"] = "egl"
+from utils.vis import render_mesh, save_obj
+cudnn.benchmark = True
+
+def load_image(image_path):
+    # load image
+    image_pil = Image.open(image_path).convert("RGB")  # load image
+
+    transform = T.Compose(
+        [
+            T.RandomResize([800], max_size=1333),
+            T.ToTensor(),
+            T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
+        ]
+    )
+    image, _ = transform(image_pil, None)  # 3, h, w
+    return image_pil, image
+
+
+def load_model(model_config_path, model_checkpoint_path, device):
+    args = SLConfig.fromfile(model_config_path)
+    args.device = device
+    model = build_model(args)
+    checkpoint = torch.load(model_checkpoint_path, map_location="cpu")
+    load_res = model.load_state_dict(clean_state_dict(checkpoint["model"]), strict=False)
+    print(load_res)
+    _ = model.eval()
+    return model
+
+
+def get_grounding_output(model, image, caption, box_threshold, text_threshold, with_logits=True, device="cpu"):
+    caption = caption.lower()
+    caption = caption.strip()
+    if not caption.endswith("."):
+        caption = caption + "."
+    model = model.to(device)
+    image = image.to(device)
+    with torch.no_grad():
+        outputs = model(image[None], captions=[caption])
+    logits = outputs["pred_logits"].cpu().sigmoid()[0]  # (nq, 256)
+    boxes = outputs["pred_boxes"].cpu()[0]  # (nq, 4)
+    logits.shape[0]
+
+    # filter output
+    logits_filt = logits.clone()
+    boxes_filt = boxes.clone()
+    filt_mask = logits_filt.max(dim=1)[0] > box_threshold
+    logits_filt = logits_filt[filt_mask]  # num_filt, 256
+    boxes_filt = boxes_filt[filt_mask]  # num_filt, 4
+    logits_filt.shape[0]
+
+    # get phrase
+    tokenlizer = model.tokenizer
+    tokenized = tokenlizer(caption)
+    # build pred
+    pred_phrases = []
+    for logit, box in zip(logits_filt, boxes_filt):
+        pred_phrase = get_phrases_from_posmap(logit > text_threshold, tokenized, tokenlizer)
+        if with_logits:
+            pred_phrases.append(pred_phrase + f"({str(logit.max().item())[:4]})")
+        else:
+            pred_phrases.append(pred_phrase)
+
+    return boxes_filt, pred_phrases
+
+
+def show_mask(mask, ax, random_color=False):
+    if random_color:
+        color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
+    else:
+        color = np.array([30 / 255, 144 / 255, 255 / 255, 0.6])
+    h, w = mask.shape[-2:]
+    mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
+    ax.imshow(mask_image)
+
+def show_box(box, ax, label):
+    x0, y0 = box[0], box[1]
+    w, h = box[2] - box[0], box[3] - box[1]
+    if 'person' in label.lower() or 'human' in label.lower():
+        color = 'green'
+    else:
+        color = 'blue'
+    ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor=color, facecolor=(0, 0, 0, 0), lw=2))
+    ax.text(x0, y0-5, label, fontsize=5, color='white',bbox={'facecolor': color, 'alpha': 0.7, 'pad': 1, 'edgecolor': 'none'})
+
+def save_mask_data(output_dir, mask_list, box_list, label_list):
+    value = 0  # 0 for background
+
+    mask_img = torch.zeros(mask_list.shape[-2:])
+    for idx, mask in enumerate(mask_list):
+        mask_img[mask.cpu().numpy()[0] == True] = value + idx + 1
+    plt.figure(figsize=(10, 10))
+    plt.imshow(mask_img.numpy())
+    plt.axis('off')
+    plt.savefig(os.path.join(output_dir, 'mask.jpg'), bbox_inches="tight", dpi=300, pad_inches=0.0)
+
+    json_data = [{
+        'value': value,
+        'label': 'background'
+    }]
+    for label, box in zip(label_list, box_list):
+        value += 1
+        name, logit = label.split('(')
+        logit = logit[:-1]  # the last is ')'
+        json_data.append({
+            'value': value,
+            'label': name,
+            'logit': float(logit),
+            'box': box.numpy().tolist(),
+        })
+    with open(os.path.join(output_dir, 'mask.json'), 'w') as f:
+        json.dump(json_data, f)
+
+def bbox_resize(bbox, scale=1.0):
+    center = (bbox[2:] + bbox[:2]) / 2
+    new_size = (bbox[2:] - bbox[:2]) * scale
+    new_bbox = torch.cat((center - new_size / 2, center + new_size / 2))
+    return new_bbox
+
+def mesh_recovery(original_img, bboxes):
+    transform = transforms.ToTensor()
+    original_img_height, original_img_width = original_img.shape[:2]
+
+    vis_img = original_img.copy()
+    for bbox in bboxes:  # [x1, y1, x2, y2]
+        bbox = [bbox[0], bbox[1], bbox[2] - bbox[0], bbox[3] - bbox[1]]   # xyxy -> xyhw
+        bbox = process_bbox(bbox, original_img_width, original_img_height)
+        img, img2bb_trans, bb2img_trans = generate_patch_image(original_img, bbox, 1.0, 0.0, False, cfg.input_img_shape)
+        img = transform(img.astype(np.float32)) / 255
+        img = img.cuda()[None, :, :, :]
+
+        # forward
+        inputs = {'img': img}
+        with torch.no_grad():
+            out = model(inputs, 'test')
+        mesh = out['smplx_mesh_cam'].detach().cpu().numpy()[0]
+
+        # # save mesh
+        # save_obj(mesh, smpl_x.face, 'output.obj')
+
+        focal = [cfg.focal[0] / cfg.input_body_shape[1] * bbox[2], cfg.focal[1] / cfg.input_body_shape[0] * bbox[3]]
+        princpt = [cfg.princpt[0] / cfg.input_body_shape[1] * bbox[2] + bbox[0],
+                   cfg.princpt[1] / cfg.input_body_shape[0] * bbox[3] + bbox[1]]
+        rendered_img, _ = render_mesh(vis_img[:, :, ::-1], mesh, smpl_x.face, {'focal': focal, 'princpt': princpt})
+        vis_img = rendered_img.copy()
+
+    return rendered_img
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser("Grounded-Segment-Anything Demo", add_help=True)
+    parser.add_argument("--config", type=str, required=True, help="path to config file")
+    parser.add_argument(
+        "--grounded_checkpoint", type=str, required=True, help="path to checkpoint file"
+    )
+    parser.add_argument(
+        "--sam_checkpoint", type=str, required=True, help="path to checkpoint file"
+    )
+    parser.add_argument(
+        "--osx_checkpoint", type=str, required=True, help="path to checkpoint file"
+    )
+    parser.add_argument("--input_image", type=str, required=True, help="path to image file")
+    parser.add_argument("--text_prompt", type=str, required=True, help="text prompt")
+    parser.add_argument(
+        "--output_dir", "-o", type=str, default="outputs", required=True, help="output directory"
+    )
+
+    parser.add_argument("--box_threshold", type=float, default=0.3, help="box threshold")
+    parser.add_argument("--text_threshold", type=float, default=0.25, help="text threshold")
+
+    parser.add_argument("--device", type=str, default="cpu", help="running on cpu only!, default=False")
+    args = parser.parse_args()
+
+    # cfg
+    config_file = args.config  # change the path of the model config file
+    grounded_checkpoint = args.grounded_checkpoint  # change the path of the model
+    sam_checkpoint = args.sam_checkpoint
+    osx_checkpoint = args.osx_checkpoint
+    image_path = args.input_image
+    text_prompt = args.text_prompt
+    output_dir = args.output_dir
+    box_threshold = args.box_threshold
+    text_threshold = args.text_threshold
+    device = args.device
+
+    # make dir
+    os.makedirs(output_dir, exist_ok=True)
+    # load image
+    image_pil, image = load_image(image_path)
+    # load model
+    model = load_model(config_file, grounded_checkpoint, device=device)
+
+    # visualize raw image
+    image_pil.save(os.path.join(output_dir, "raw_image.jpg"))
+
+    # run grounding dino model
+    boxes_filt, pred_phrases = get_grounding_output(
+        model, image, text_prompt, box_threshold, text_threshold, device=device
+    )
+
+    # initialize SAM
+    sam = build_sam(checkpoint=sam_checkpoint)
+    sam.to(device=device)
+    predictor = SamPredictor(sam)
+    image = cv2.imread(image_path)
+    image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+    predictor.set_image(image)
+
+    # initialize OSX
+    model = get_model()
+    model = DataParallel(model).cuda()
+    ckpt = torch.load(osx_checkpoint)
+    model.load_state_dict(ckpt['network'], strict=False)
+    model.eval()
+
+    size = image_pil.size
+    H, W = size[1], size[0]
+    for i in range(boxes_filt.size(0)):
+        boxes_filt[i] = boxes_filt[i] * torch.Tensor([W, H, W, H])
+        boxes_filt[i][:2] -= boxes_filt[i][2:] / 2
+        boxes_filt[i][2:] += boxes_filt[i][:2]
+
+    boxes_filt = boxes_filt.cpu()
+    transformed_boxes = predictor.transform.apply_boxes_torch(boxes_filt, image.shape[:2]).to(device)
+
+    masks, _, _ = predictor.predict_torch(
+        point_coords=None,
+        point_labels=None,
+        boxes=transformed_boxes,
+        multimask_output=False,
+    )
+
+    # scale up the human bboxes
+    boxes_human = []
+    for i, label in enumerate(pred_phrases):
+        if 'person' in label.lower() or 'human' in label.lower():
+            boxes_filt[i] = bbox_resize(boxes_filt[i], scale=1.1)
+            boxes_human.append(boxes_filt[i])
+
+    # predict and visualize 3d human mesh
+    for i, label in enumerate(pred_phrases):
+        if 'person' in label.lower() or 'man' in label.lower():
+            boxes_human.append(boxes_filt[i])
+    rendered_img = mesh_recovery(image, boxes_human)
+    cv2.imwrite(os.path.join(output_dir, "grounded_sam_osx_output.jpg"), rendered_img)
+
+    # draw output image
+    fig, (plt1, plt2) = plt.subplots(ncols=2, figsize=(10, 20), gridspec_kw={'wspace':0, 'hspace':0})
+
+    plt1.imshow(image)
+    for mask in masks:
+        show_mask(mask.cpu().numpy(), plt1, random_color=True)
+    for box, label in zip(boxes_filt, pred_phrases):
+        show_box(box.numpy(), plt1, label)
+    rendered_img = cv2.imread(os.path.join(output_dir, "grounded_sam_osx_output.jpg"))
+    plt2.imshow(rendered_img)
+    for box, label in zip(boxes_filt, pred_phrases):
+        show_box(box.numpy(), plt2, label)
+    plt1.axis('off')
+    plt2.axis('off')
+    plt.savefig(
+        os.path.join(output_dir, "grounded_sam_osx_output.jpg"),
+        bbox_inches="tight", dpi=300, pad_inches=0.0
+    )
+
+    save_mask_data(output_dir, masks, boxes_filt, pred_phrases)
+
+

external/Grounded-Segment-Anything/grounding_dino_demo.py

@@ -0,0 +1,31 @@
+from groundingdino.util.inference import load_model, load_image, predict, annotate, Model
+import cv2
+
+
+CONFIG_PATH = "GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py"
+CHECKPOINT_PATH = "./groundingdino_swint_ogc.pth"
+DEVICE = "cuda"
+IMAGE_PATH = "assets/demo7.jpg"
+TEXT_PROMPT = "Horse. Clouds. Grasses. Sky. Hill."
+BOX_TRESHOLD = 0.35
+TEXT_TRESHOLD = 0.25
+FP16_INFERENCE = True
+
+image_source, image = load_image(IMAGE_PATH)
+model = load_model(CONFIG_PATH, CHECKPOINT_PATH)
+
+if FP16_INFERENCE:
+    image = image.half()
+    model = model.half()
+
+boxes, logits, phrases = predict(
+    model=model,
+    image=image,
+    caption=TEXT_PROMPT,
+    box_threshold=BOX_TRESHOLD,
+    text_threshold=TEXT_TRESHOLD,
+    device=DEVICE,
+)
+
+annotated_frame = annotate(image_source=image_source, boxes=boxes, logits=logits, phrases=phrases)
+cv2.imwrite("annotated_image.jpg", annotated_frame)

external/Grounded-Segment-Anything/predict.py

@@ -0,0 +1,288 @@
+# Prediction interface for Cog ⚙️
+# https://github.com/replicate/cog/blob/main/docs/python.md
+
+import os
+import json
+from typing import Any
+import numpy as np
+import random
+import torch
+import torchvision
+import torchvision.transforms as transforms
+from PIL import Image
+import cv2
+import matplotlib.pyplot as plt
+from cog import BasePredictor, Input, Path, BaseModel
+
+from subprocess import call
+
+HOME = os.getcwd()
+os.chdir("GroundingDINO")
+call("pip install -q .", shell=True)
+os.chdir(HOME)
+os.chdir("segment_anything")
+call("pip install -q .", shell=True)
+os.chdir(HOME)
+
+# Grounding DINO
+import GroundingDINO.groundingdino.datasets.transforms as T
+from GroundingDINO.groundingdino.models import build_model
+from GroundingDINO.groundingdino.util.slconfig import SLConfig
+from GroundingDINO.groundingdino.util.utils import (
+    clean_state_dict,
+    get_phrases_from_posmap,
+)
+
+# segment anything
+from segment_anything import build_sam, build_sam_hq, SamPredictor
+
+from ram.models import ram
+
+
+class ModelOutput(BaseModel):
+    tags: str
+    rounding_box_img: Path
+    masked_img: Path
+    json_data: Any
+
+
+class Predictor(BasePredictor):
+    def setup(self):
+        """Load the model into memory to make running multiple predictions efficient"""
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        normalize = transforms.Normalize(
+            mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
+        )
+        self.image_size = 384
+        self.transform = transforms.Compose(
+            [
+                transforms.Resize((self.image_size, self.image_size)),
+                transforms.ToTensor(),
+                normalize,
+            ]
+        )
+
+        # load model
+        self.ram_model = ram(
+            pretrained="pretrained/ram_swin_large_14m.pth",
+            image_size=self.image_size,
+            vit="swin_l",
+        )
+        self.ram_model.eval()
+        self.ram_model = self.ram_model.to(self.device)
+
+        self.model = load_model(
+            "GroundingDINO/groundingdino/config/GroundingDINO_SwinT_OGC.py",
+            "pretrained/groundingdino_swint_ogc.pth",
+            device=self.device,
+        )
+
+        self.sam = SamPredictor(
+            build_sam(checkpoint="pretrained/sam_vit_h_4b8939.pth").to(self.device)
+        )
+        self.sam_hq = SamPredictor(
+            build_sam_hq(checkpoint="pretrained/sam_hq_vit_h.pth").to(self.device)
+        )
+
+    def predict(
+        self,
+        input_image: Path = Input(description="Input image"),
+        use_sam_hq: bool = Input(
+            description="Use sam_hq instead of SAM for prediction", default=False
+        ),
+    ) -> ModelOutput:
+        """Run a single prediction on the model"""
+
+        # default settings
+        box_threshold = 0.25
+        text_threshold = 0.2
+        iou_threshold = 0.5
+
+        image_pil, image = load_image(str(input_image))
+
+        raw_image = image_pil.resize((self.image_size, self.image_size))
+        raw_image = self.transform(raw_image).unsqueeze(0).to(self.device)
+
+        with torch.no_grad():
+            tags, tags_chinese = self.ram_model.generate_tag(raw_image)
+
+        tags = tags[0].replace(" |", ",")
+
+        # run grounding dino model
+        boxes_filt, scores, pred_phrases = get_grounding_output(
+            self.model, image, tags, box_threshold, text_threshold, device=self.device
+        )
+
+        predictor = self.sam_hq if use_sam_hq else self.sam
+
+        image = cv2.imread(str(input_image))
+        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+        predictor.set_image(image)
+
+        size = image_pil.size
+        H, W = size[1], size[0]
+        for i in range(boxes_filt.size(0)):
+            boxes_filt[i] = boxes_filt[i] * torch.Tensor([W, H, W, H])
+            boxes_filt[i][:2] -= boxes_filt[i][2:] / 2
+            boxes_filt[i][2:] += boxes_filt[i][:2]
+
+        boxes_filt = boxes_filt.cpu()
+        # use NMS to handle overlapped boxes
+        print(f"Before NMS: {boxes_filt.shape[0]} boxes")
+        nms_idx = (
+            torchvision.ops.nms(boxes_filt, scores, iou_threshold).numpy().tolist()
+        )
+        boxes_filt = boxes_filt[nms_idx]
+        pred_phrases = [pred_phrases[idx] for idx in nms_idx]
+        print(f"After NMS: {boxes_filt.shape[0]} boxes")
+
+        transformed_boxes = predictor.transform.apply_boxes_torch(
+            boxes_filt, image.shape[:2]
+        ).to(self.device)
+
+        masks, _, _ = predictor.predict_torch(
+            point_coords=None,
+            point_labels=None,
+            boxes=transformed_boxes.to(self.device),
+            multimask_output=False,
+        )
+
+        # draw output image
+        plt.figure(figsize=(10, 10))
+        for mask in masks:
+            show_mask(mask.cpu().numpy(), plt.gca(), random_color=True)
+        for box, label in zip(boxes_filt, pred_phrases):
+            show_box(box.numpy(), plt.gca(), label)
+
+        rounding_box_path = "/tmp/automatic_label_output.png"
+        plt.axis("off")
+        plt.savefig(
+            Path(rounding_box_path), bbox_inches="tight", dpi=300, pad_inches=0.0
+        )
+        plt.close()
+
+        # save masks and json data
+        value = 0  # 0 for background
+        mask_img = torch.zeros(masks.shape[-2:])
+        for idx, mask in enumerate(masks):
+            mask_img[mask.cpu().numpy()[0] == True] = value + idx + 1
+        plt.figure(figsize=(10, 10))
+        plt.imshow(mask_img.numpy())
+        plt.axis("off")
+        masks_path = "/tmp/mask.png"
+        plt.savefig(masks_path, bbox_inches="tight", dpi=300, pad_inches=0.0)
+        plt.close()
+
+        json_data = {
+            "tags": tags,
+            "mask": [{"value": value, "label": "background"}],
+        }
+        for label, box in zip(pred_phrases, boxes_filt):
+            value += 1
+            name, logit = label.split("(")
+            logit = logit[:-1]  # the last is ')'
+            json_data["mask"].append(
+                {
+                    "value": value,
+                    "label": name,
+                    "logit": float(logit),
+                    "box": box.numpy().tolist(),
+                }
+            )
+
+        json_path = "/tmp/label.json"
+        with open(json_path, "w") as f:
+            json.dump(json_data, f)
+
+        return ModelOutput(
+            tags=tags,
+            masked_img=Path(masks_path),
+            rounding_box_img=Path(rounding_box_path),
+            json_data=Path(json_path),
+        )
+
+
+def get_grounding_output(
+    model, image, caption, box_threshold, text_threshold, device="cpu"
+):
+    caption = caption.lower()
+    caption = caption.strip()
+    if not caption.endswith("."):
+        caption = caption + "."
+    model = model.to(device)
+    image = image.to(device)
+    with torch.no_grad():
+        outputs = model(image[None], captions=[caption])
+    logits = outputs["pred_logits"].cpu().sigmoid()[0]  # (nq, 256)
+    boxes = outputs["pred_boxes"].cpu()[0]  # (nq, 4)
+    logits.shape[0]
+
+    # filter output
+    logits_filt = logits.clone()
+    boxes_filt = boxes.clone()
+    filt_mask = logits_filt.max(dim=1)[0] > box_threshold
+    logits_filt = logits_filt[filt_mask]  # num_filt, 256
+    boxes_filt = boxes_filt[filt_mask]  # num_filt, 4
+    logits_filt.shape[0]
+
+    # get phrase
+    tokenlizer = model.tokenizer
+    tokenized = tokenlizer(caption)
+    # build pred
+    pred_phrases = []
+    scores = []
+    for logit, box in zip(logits_filt, boxes_filt):
+        pred_phrase = get_phrases_from_posmap(
+            logit > text_threshold, tokenized, tokenlizer
+        )
+        pred_phrases.append(pred_phrase + f"({str(logit.max().item())[:4]})")
+        scores.append(logit.max().item())
+
+    return boxes_filt, torch.Tensor(scores), pred_phrases
+
+
+def load_image(image_path):
+    # load image
+    image_pil = Image.open(image_path).convert("RGB")  # load image
+
+    transform = T.Compose(
+        [
+            T.RandomResize([800], max_size=1333),
+            T.ToTensor(),
+            T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
+        ]
+    )
+    image, _ = transform(image_pil, None)  # 3, h, w
+    return image_pil, image
+
+
+def load_model(model_config_path, model_checkpoint_path, device):
+    args = SLConfig.fromfile(model_config_path)
+    args.device = device
+    model = build_model(args)
+    checkpoint = torch.load(model_checkpoint_path, map_location="cpu")
+    load_res = model.load_state_dict(
+        clean_state_dict(checkpoint["model"]), strict=False
+    )
+    print(load_res)
+    _ = model.eval()
+    return model
+
+
+def show_mask(mask, ax, random_color=False):
+    if random_color:
+        color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
+    else:
+        color = np.array([30 / 255, 144 / 255, 255 / 255, 0.6])
+    h, w = mask.shape[-2:]
+    mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
+    ax.imshow(mask_image)
+
+
+def show_box(box, ax, label):
+    x0, y0 = box[0], box[1]
+    w, h = box[2] - box[0], box[3] - box[1]
+    ax.add_patch(
+        plt.Rectangle((x0, y0), w, h, edgecolor="green", facecolor=(0, 0, 0, 0), lw=1.5)
+    )
+    ax.text(x0, y0, label)

external/Grounded-Segment-Anything/requirements.txt

@@ -0,0 +1,23 @@
+addict
+diffusers
+gradio
+huggingface_hub
+matplotlib
+numpy
+onnxruntime
+opencv_python
+Pillow
+pycocotools
+PyYAML
+requests
+setuptools
+supervision
+termcolor
+timm
+torch
+torchvision
+transformers
+yapf
+nltk
+fairscale
+litellm

external/Metric3D/.gitignore

@@ -0,0 +1,5 @@
+work_dirs
+show_dirs
+*__pycache__*
+test_dataloader_*
+*.onnx

external/Metric3D/LICENSE

@@ -0,0 +1,24 @@
+BSD 2-Clause License
+
+Copyright (c) 2024, Wei Yin and Mu Hu
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

external/Metric3D/README.md

@@ -0,0 +1,396 @@
+# 🚀 Metric3D Project 🚀
+
+**Official PyTorch implementation of Metric3Dv1 and Metric3Dv2:**   
+
+[1] [Metric3D: Towards Zero-shot Metric 3D Prediction from A Single Image](https://arxiv.org/abs/2307.10984)  
+
+[2] [Metric3Dv2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth and Surface Normal Estimation](https://arxiv.org/abs/2404.15506)
+
+<a href='https://jugghm.github.io/Metric3Dv2'><img src='https://img.shields.io/badge/project%20page-@Metric3D-yellow.svg'></a>
+<a href='https://arxiv.org/abs/2307.10984'><img src='https://img.shields.io/badge/arxiv-@Metric3Dv1-green'></a>
+<a href='https://arxiv.org/abs/2404.15506'><img src='https://img.shields.io/badge/arxiv-@Metric3Dv2-red'></a>
+<a href='https://huggingface.co/spaces/JUGGHM/Metric3D'><img src='https://img.shields.io/badge/%F0%9F%A4%97%20Hugging%20Face-Spaces-blue'></a>
+
+[//]: # (### [Project Page]&#40;https://arxiv.org/abs/2307.08695&#41; | [v2 Paper]&#40;https://arxiv.org/abs/2307.10984&#41; | [v1 Arxiv]&#40;https://arxiv.org/abs/2307.10984&#41; | [Video]&#40;https://www.youtube.com/playlist?list=PLEuyXJsWqUNd04nwfm9gFBw5FVbcaQPl3&#41; | [Hugging Face 🤗]&#40;https://huggingface.co/spaces/JUGGHM/Metric3D&#41; )
+
+[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/metric3d-v2-a-versatile-monocular-geometric-1/monocular-depth-estimation-on-nyu-depth-v2)](https://paperswithcode.com/sota/monocular-depth-estimation-on-nyu-depth-v2?p=metric3d-v2-a-versatile-monocular-geometric-1) 
+
+[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/metric3d-v2-a-versatile-monocular-geometric-1/monocular-depth-estimation-on-kitti-eigen)](https://paperswithcode.com/sota/monocular-depth-estimation-on-kitti-eigen?p=metric3d-v2-a-versatile-monocular-geometric-1) 
+
+[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/metric3d-v2-a-versatile-monocular-geometric-1/surface-normals-estimation-on-nyu-depth-v2-1)](https://paperswithcode.com/sota/surface-normals-estimation-on-nyu-depth-v2-1?p=metric3d-v2-a-versatile-monocular-geometric-1) 
+
+[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/metric3d-v2-a-versatile-monocular-geometric-1/surface-normals-estimation-on-ibims-1)](https://paperswithcode.com/sota/surface-normals-estimation-on-ibims-1?p=metric3d-v2-a-versatile-monocular-geometric-1) 
+
+[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/metric3d-v2-a-versatile-monocular-geometric-1/surface-normals-estimation-on-scannetv2)](https://paperswithcode.com/sota/surface-normals-estimation-on-scannetv2?p=metric3d-v2-a-versatile-monocular-geometric-1) 
+
+🏆 **Champion in [CVPR2023 Monocular Depth Estimation Challenge](https://jspenmar.github.io/MDEC)**
+
+## News 
+- `[2024/8]` Metric3Dv2 is accepted by TPAMI!
+- `[2024/7/5]` Our stable-diffusion alternative GeoWizard has now been accepted by ECCV 2024! Check NOW the [repository](https://github.com/fuxiao0719/GeoWizard) and [paper](https://arxiv.org/abs/2403.12013) for the finest-grained geometry ever! 🎉🎉🎉
+- `[2024/6/25]` Json files for KITTI datasets now available! Refer to [Training](./training/README.md) for more details 
+- `[2024/6/3]` ONNX is supported! We appreciate [@xenova](https://github.com/xenova) for their remarkable efforts! 
+- `[2024/4/25]` Weights for ViT-giant2 model released!  
+- `[2024/4/11]` Training codes are released!
+- `[2024/3/18]` [HuggingFace 🤗](https://huggingface.co/spaces/JUGGHM/Metric3D) GPU version updated!
+- `[2024/3/18]` [Project page](https://jugghm.github.io/Metric3Dv2/) released!
+- `[2024/3/18]` Metric3D V2 models released, supporting metric depth and surface normal now!
+- `[2023/8/10]` Inference codes, pre-trained weights, and demo released.
+- `[2023/7]` Metric3D accepted by ICCV 2023!
+- `[2023/4]` The Champion of [2nd Monocular Depth Estimation Challenge](https://jspenmar.github.io/MDEC) in CVPR 2023
+
+##  🌼 Abstract
+Metric3D is a strong and robust geometry foundation model for high-quality and zero-shot **metric depth** and **surface normal** estimation from a single image. It excels at solving in-the-wild scene reconstruction. It can directly help you measure the size of structures from a single image. Now it achieves SOTA performance on over 10 depth and normal benchmarks.
+
+![depth_normal](media/screenshots/depth_normal.jpg)
+
+![metrology](media/screenshots/metrology.jpg)
+
+
+##  📝 Benchmarks 
+
+### Metric Depth
+
+[//]: # (#### Zero-shot Testing)
+
+[//]: # (Our models work well on both indoor and outdoor scenarios, compared with other zero-shot metric depth estimation methods.)
+
+[//]: # ()
+[//]: # (|                 | Backbone   | KITTI $\delta 1$ ↑ | KITTI $\delta 2$  ↑ | KITTI $\delta 3$ ↑ | KITTI AbsRel  ↓ | KITTI RMSE  ↓ | KITTI RMS_log  ↓ | NYU $\delta 1$ ↑ | NYU $\delta 2$ ↑ | NYU $\delta 3$ ↑ | NYU AbsRel  ↓ | NYU RMSE  ↓ | NYU log10  ↓ |)
+
+[//]: # (|-----------------|------------|--------------------|---------------------|--------------------|-----------------|---------------|------------------|------------------|------------------|------------------|---------------|-------------|--------------|)
+
+[//]: # (| ZeroDepth       | ResNet-18 | 0.910              | 0.980               | 0.996              | 0.057           | 4.044         | 0.083            | 0.901            | 0.961            | -                | 0.100         | 0.380       | -            |)
+
+[//]: # (| PolyMax         | ConvNeXt-L    | -                  | -                   | -                  | -               | -             | -                | 0.969            | 0.996            | 0.999            | 0.067         | 0.250       | 0.033        |)
+
+[//]: # (| Ours | ViT-L     | 0.985              | 0.995               | 0.999              | 0.052           | 2.511         | 0.074            | 0.975            | 0.994            | 0.998            | 0.063         | 0.251       | 0.028        |)
+
+[//]: # (| Ours | ViT-g2    | 0.989              | 0.996               | 0.999              | 0.051           | 2.403         | 0.080            | 0.980            | 0.997            | 0.999            | 0.067         | 0.260       | 0.030        |)
+
+[//]: # ()
+[//]: # ([//]: # &#40;| Adabins | Efficient-B5 | 0.964 | 0.995 | 0.999 | 0.058  |  2.360 | 0.088            | 0.903  | 0.984  | 0.997  | 0.103  | 0.0444  | 0.364 |&#41;)
+[//]: # ([//]: # &#40;| NewCRFs | SwinT-L | 0.974 | 0.997 | 0.999 | 0.052  |  2.129 | 0.079            | 0.922  | 0.983  | 0.994  | 0.095  | 0.041  | 0.334 |&#41;)
+[//]: # ([//]: # &#40;| Ours &#40;CSTM_label&#41; | ConvNeXt-L |      0.964      | 0.993   | 0.998  | 0.058 | 2.770  | 0.092            | 0.944  |  0.986 | 0.995   | 0.083  |  0.035 |  0.310 |&#41;)
+
+[//]: # (#### Finetuned)
+Our models rank 1st on the routing KITTI and NYU benchmarks.
+
+|               | Backbone    | KITTI δ1 ↑ | KITTI δ2  ↑  | KITTI AbsRel  ↓ | KITTI RMSE  ↓ | KITTI RMS_log  ↓ | NYU δ1 ↑ | NYU δ2 ↑  | NYU AbsRel  ↓ | NYU RMSE  ↓ | NYU log10  ↓ |
+|---------------|-------------|------------|-------------|-----------------|---------------|------------------|----------|----------|---------------|-------------|--------------|
+| ZoeDepth      | ViT-Large   | 0.971      | 0.995                  | 0.053           | 2.281         | 0.082            | 0.953    | 0.995        | 0.077         | 0.277       | 0.033        |
+| ZeroDepth     | ResNet-18   | 0.968      | 0.996                   | 0.057           | 2.087         | 0.083            | 0.954    | 0.995           | 0.074         | 0.269       | 0.103        |
+| IEBins        | SwinT-Large | 0.978      | 0.998                  | 0.050           | 2.011         | 0.075            | 0.936    | 0.992           | 0.087         | 0.314       | 0.031        |
+| DepthAnything | ViT-Large   | 0.982      | 0.998                  | 0.046           | 1.985         | 0.069            | 0.984    | 0.998           | 0.056         | 0.206       | 0.024        |
+| Ours          | ViT-Large   | 0.985      | 0.998       | 0.044                        | 1.985         | 0.064            | 0.989    | 0.998           | 0.047         | 0.183       | 0.020        |
+| Ours          | ViT-giant2  | 0.989      | 0.998       | 0.039                        | 1.766         | 0.060            | 0.987    | 0.997           | 0.045         | 0.187       | 0.015        |
+
+### Affine-invariant Depth
+Even compared to recent affine-invariant depth methods (Marigold and Depth Anything), our metric-depth (and normal) models still show superior performance. 
+
+|                       | #Data for Pretrain and Train                 | KITTI Absrel ↓ | KITTI δ1 ↑ | NYUv2 AbsRel  ↓ | NYUv2 δ1 ↑ | DIODE-Full AbsRel ↓ | DIODE-Full δ1 ↑ | Eth3d AbsRel  ↓ | Eth3d δ1 ↑ |
+|-----------------------|----------------------------------------------|----------------|------------|-----------------|------------|---------------------|-----------------|----------------------|------------|
+| OmniData (v2, ViT-L)       | 1.3M + 12.2M                                 | 0.069          | 0.948      | 0.074           | 0.945      | 0.149               | 0.835           | 0.166                | 0.778      | 
+| MariGold  (LDMv2)     | 5B + 74K                                     | 0.099          | 0.916      | 0.055           | 0.961      | 0.308               | 0.773           | 0.127                | 0.960      | 
+| DepthAnything (ViT-L) | 142M + 63M                                   | 0.076          | 0.947      | 0.043           | 0.981      | 0.277               | 0.759           | 0.065                | 0.882      | 
+| Ours (ViT-L)          | 142M + 16M                                   | 0.042          | 0.979      | 0.042           | 0.980      | 0.141               | 0.882           | 0.042                | 0.987      | 
+| Ours (ViT-g)          | 142M + 16M                                   | 0.043          | 0.982      | 0.043           | 0.981      | 0.136               | 0.895           | 0.042                | 0.983      | 
+
+
+### Surface Normal
+Our models also show powerful performance on normal benchmarks.
+
+|              | NYU 11.25° ↑ | NYU Mean ↓ | NYU RMS ↓ | ScanNet 11.25° ↑ | ScanNet Mean ↓ | ScanNet RMS ↓ | iBims 11.25° ↑ | iBims Mean ↓ | iBims RMS ↓ | 
+|--------------|----------|----------|-----------|-----------------|----------------|--------------|---------------|--------------|-------------|
+| EESNU        | 0.597    | 16.0     | 24.7      | 0.711           | 11.8           | 20.3         | 0.585         | 20.0         | -           | 
+| IronDepth    | -        | -        | -         | -               | -              | -            | 0.431         | 25.3         | 37.4        | 
+| PolyMax      | 0.656    | 13.1     | 20.4      | -               | -              | -            | -             | -            | -           |
+| Ours (ViT-L) | 0.688    | 12.0     | 19.2      | 0.760           | 9.9            | 16.4         | 0.694         | 19.4         | 34.9        | 
+| Ours (ViT-g)   | 0.662    | 13.2     | 20.2      | 0.778           | 9.2            | 15.3         | 0.697         | 19.6         | 35.2        |
+
+
+
+## 🌈 DEMOs
+
+### Zero-shot monocular metric depth & surface normal
+<img src="media/gifs/demo_1.gif" width="600" height="337">  
+<img src="media/gifs/demo_12.gif" width="600" height="337">
+
+### Zero-shot metric 3D recovery
+<img src="media/gifs/demo_2.gif" width="600" height="337">  
+
+### Improving monocular SLAM
+<img src="media/gifs/demo_22.gif" width="600" height="337">  
+
+[//]: # (https://github.com/YvanYin/Metric3D/assets/35299633/f95815ef-2506-4193-a6d9-1163ea821268)
+
+[//]: # (https://github.com/YvanYin/Metric3D/assets/35299633/ed00706c-41cc-49ea-accb-ad0532633cc2)
+
+[//]: # (### Zero-shot metric 3D recovery)
+
+[//]: # (https://github.com/YvanYin/Metric3D/assets/35299633/26cd7ae1-dd5a-4446-b275-54c5ca7ef945)
+
+[//]: # (https://github.com/YvanYin/Metric3D/assets/35299633/21e5484b-c304-4fe3-b1d3-8eebc4e26e42)
+[//]: # (### Monocular reconstruction for a Sequence)
+
+[//]: # ()
+[//]: # (### In-the-wild 3D reconstruction)
+
+[//]: # ()
+[//]: # (|           | Image | Reconstruction | Pointcloud File |)
+
+[//]: # (|:---------:|:------------------:|:------------------:|:--------:|)
+
+[//]: # (|    room   |    <img src="data/wild_demo/jonathan-borba-CnthDZXCdoY-unsplash.jpg" width="300" height="335">     |     <img src="media/gifs/room.gif" width="300" height="335">            |  [Download]&#40;https://drive.google.com/file/d/1P1izSegH2c4LUrXGiUksw037PVb0hjZr/view?usp=drive_link&#41;        |)
+
+[//]: # (| Colosseum |    <img src="data/wild_demo/david-kohler-VFRTXGw1VjU-unsplash.jpg" width="300" height="169">     |     <img src="media/gifs/colo.gif" width="300" height="169">         |     [Download]&#40;https://drive.google.com/file/d/1jJCXe5IpxBhHDr0TZtNZhjxKTRUz56Hg/view?usp=drive_link&#41;     |)
+
+[//]: # (|   chess   |    <img src="data/wild_demo/randy-fath-G1yhU1Ej-9A-unsplash.jpg" width="300" height="169" align=center>     |     <img src="media/gifs/chess.gif" width="300" height="169">            |      [Download]&#40;https://drive.google.com/file/d/1oV_Foq25_p-tTDRTcyO2AzXEdFJQz-Wm/view?usp=drive_link&#41;    |)
+
+[//]: # ()
+[//]: # (All three images are downloaded from [unplash]&#40;https://unsplash.com/&#41; and put in the data/wild_demo directory.)
+
+[//]: # ()
+[//]: # (### 3D metric reconstruction, Metric3D × DroidSLAM)
+
+[//]: # (Metric3D can also provide scale information for DroidSLAM, help to solve the scale drift problem for better trajectories. )
+
+[//]: # ()
+[//]: # (#### Bird Eyes' View &#40;Left: Droid-SLAM &#40;mono&#41;. Right: Droid-SLAM with Metric-3D&#41;)
+
+[//]: # ()
+[//]: # (<div align=center>)
+
+[//]: # (<img src="media/gifs/0028.gif"> )
+
+[//]: # (</div>)
+
+[//]: # ()
+[//]: # (### Front View)
+
+[//]: # ()
+[//]: # (<div align=center>)
+
+[//]: # (<img src="media/gifs/0028_fv.gif"> )
+
+[//]: # (</div>)
+
+[//]: # ()
+[//]: # (#### KITTI odemetry evaluation &#40;Translational RMS drift &#40;t_rel, ↓&#41; / Rotational RMS drift &#40;r_rel, ↓&#41;&#41;)
+
+[//]: # (|            | Modality |   seq 00   |   seq 02   |   seq 05  |   seq 06   |   seq 08   |   seq 09  |   seq 10  |)
+
+[//]: # (|:----------:|:--------:|:----------:|:----------:|:---------:|:----------:|:----------:|:---------:|:---------:|)
+
+[//]: # (|  ORB-SLAM2 |   Mono   | 11.43/0.58 | 10.34/0.26 | 9.04/0.26 | 14.56/0.26 | 11.46/0.28 |  9.3/0.26 | 2.57/0.32 |)
+
+[//]: # (| Droid-SLAM |   Mono   |  33.9/0.29 | 34.88/0.27 | 23.4/0.27 |  17.2/0.26 |  39.6/0.31 | 21.7/0.23 |   7/0.25  |)
+
+[//]: # (| Droid+Ours |   Mono   |  1.44/0.37 |  2.64/0.29 | 1.44/0.25 |   0.6/0.2  |   2.2/0.3  | 1.63/0.22 | 2.73/0.23 |)
+
+[//]: # (|  ORB-SLAM2 |  Stereo  |  0.88/0.31 |  0.77/0.28 | 0.62/0.26 |  0.89/0.27 |  1.03/0.31 | 0.86/0.25 | 0.62/0.29 |)
+
+[//]: # ()
+[//]: # (Metric3D makes the mono-SLAM scale-aware, like stereo systems.)
+
+[//]: # ()
+[//]: # (#### KITTI sequence videos - Youtube)
+
+[//]: # ([2011_09_30_drive_0028]&#40;https://youtu.be/gcTB4MgVCLQ&#41; /)
+
+[//]: # ([2011_09_30_drive_0033]&#40;https://youtu.be/He581fmoPP4&#41; /)
+
+[//]: # ([2011_09_30_drive_0034]&#40;https://youtu.be/I3PkukQ3_F8&#41;)
+
+[//]: # ()
+[//]: # (#### Estimated pose)
+
+[//]: # ([2011_09_30_drive_0033]&#40;https://drive.google.com/file/d/1SMXWzLYrEdmBe6uYMR9ShtDXeFDewChv/view?usp=drive_link&#41; / )
+
+[//]: # ([2011_09_30_drive_0034]&#40;https://drive.google.com/file/d/1ONU4GxpvTlgW0TjReF1R2i-WFxbbjQPG/view?usp=drive_link&#41; /)
+
+[//]: # ([2011_10_03_drive_0042]&#40;https://drive.google.com/file/d/19fweg6p1Q6TjJD2KlD7EMA_aV4FIeQUD/view?usp=drive_link&#41;)
+
+[//]: # ()
+[//]: # (#### Pointcloud files)
+
+[//]: # ([2011_09_30_drive_0033]&#40;https://drive.google.com/file/d/1K0o8DpUmLf-f_rue0OX1VaHlldpHBAfw/view?usp=drive_link&#41; /)
+
+[//]: # ([2011_09_30_drive_0034]&#40;https://drive.google.com/file/d/1bvZ6JwMRyvi07H7Z2VD_0NX1Im8qraZo/view?usp=drive_link&#41; /)
+
+[//]: # ([2011_10_03_drive_0042]&#40;https://drive.google.com/file/d/1Vw59F8nN5ApWdLeGKXvYgyS9SNKHKy4x/view?usp=drive_link&#41;)
+
+## 🔨 Installation
+### One-line Installation
+For the ViT models, use the following environment：
+```bash
+pip install -r requirements_v2.txt
+```
+
+For ConvNeXt-L, it is 
+```bash
+pip install -r requirements_v1.txt
+```
+
+### dataset annotation components
+With off-the-shelf depth datasets, we need to generate json annotaions in compatible with this dataset, which is organized by:
+```
+dict(
+	'files':list(
+		dict(
+			'rgb': 'data/kitti_demo/rgb/xxx.png',
+			'depth': 'data/kitti_demo/depth/xxx.png',
+			'depth_scale': 1000.0 # the depth scale of gt depth img.
+			'cam_in': [fx, fy, cx, cy],
+		),
+
+		dict(
+			...
+		),
+
+		...
+	)
+)
+```
+To generate such annotations, please refer to the "Inference" section.
+
+### configs
+In ```mono/configs``` we provide different config setups. 
+
+Intrinsics of the canonical camera is set bellow: 
+```
+    canonical_space = dict(
+        img_size=(512, 960),
+        focal_length=1000.0,
+    ),
+```
+where cx and cy is set to be half of the image size.
+
+Inference settings are defined as
+```
+    depth_range=(0, 1),
+    depth_normalize=(0.3, 150),
+    crop_size = (512, 1088),
+```
+where the images will be first resized as the ```crop_size``` and then fed into the model.
+
+## ✈️ Training
+Please refer to [training/README.md](./training/README.md).
+Now we provide complete json files for KITTI fine-tuning.
+
+## ✈️ Inference
+### News: Improved ONNX support with dynamic shapes (Feature owned by [@xenova](https://github.com/xenova). Appreciate for this outstanding contribution 🚩🚩🚩)
+
+Now the onnx supports are availble for all three models with varying shapes. Refer to [issue117](https://github.com/YvanYin/Metric3D/issues/117) for more details.
+
+### Improved ONNX Checkpoints Available now 
+|      |       Encoder       |      Decoder      |                                               Link                                                |
+|:----:|:-------------------:|:-----------------:|:-------------------------------------------------------------------------------------------------:|
+| v2-S-ONNX | DINO2reg-ViT-Small  |    RAFT-4iter     | [Download 🤗](https://huggingface.co/onnx-community/metric3d-vit-small) |
+| v2-L-ONNX | DINO2reg-ViT-Large  |    RAFT-8iter     | [Download 🤗](https://huggingface.co/onnx-community/metric3d-vit-large) |
+| v2-g-ONNX | DINO2reg-ViT-giant2 |    RAFT-8iter     | [Download 🤗](https://huggingface.co/onnx-community/metric3d-vit-giant2) |
+
+One additional [reminder](https://github.com/YvanYin/Metric3D/issues/143#issue-2444506808) for using these onnx models is reported by @norbertlink.
+
+### News: Pytorch Hub is supported
+Now you can use Metric3D via Pytorch Hub with just few lines of code:
+```python
+import torch
+model = torch.hub.load('yvanyin/metric3d', 'metric3d_vit_small', pretrain=True)
+pred_depth, confidence, output_dict = model.inference({'input': rgb})
+pred_normal = output_dict['prediction_normal'][:, :3, :, :] # only available for Metric3Dv2 i.e., ViT models
+normal_confidence = output_dict['prediction_normal'][:, 3, :, :] # see https://arxiv.org/abs/2109.09881 for details
+```
+Supported models: `metric3d_convnext_tiny`, `metric3d_convnext_large`, `metric3d_vit_small`, `metric3d_vit_large`, `metric3d_vit_giant2`.
+
+We also provided a minimal working example in [hubconf.py](https://github.com/YvanYin/Metric3D/blob/main/hubconf.py#L145), which hopefully makes everything clearer.
+
+### News: ONNX Exportation and Inference are supported
+
+We also provided a flexible working example in [metric3d_onnx_export.py](./onnx/metric3d_onnx_export.py) to export the Pytorch Hub model to ONNX format. We could test with the following commands:
+
+```bash
+# Export the model to ONNX model
+python3 onnx/metric_3d_onnx_export.py metric3d_vit_small # metric3d_vit_large/metric3d_convnext_large
+
+# Test the inference of the ONNX model
+python3 onnx/test_onnx.py metric3d_vit_small.onnx
+```
+
+[ros2_vision_inference](https://github.com/Owen-Liuyuxuan/ros2_vision_inference) provides a Python example, showcasing a pipeline from image to point clouds and integrated into ROS2 systems.
+
+### Download Checkpoint
+|      |       Encoder       |      Decoder      |                                               Link                                                |
+|:----:|:-------------------:|:-----------------:|:-------------------------------------------------------------------------------------------------:|
+| v1-T |    ConvNeXt-Tiny    | Hourglass-Decoder | [Download 🤗](https://huggingface.co/JUGGHM/Metric3D/blob/main/convtiny_hourglass_v1.pth)        |
+| v1-L |   ConvNeXt-Large    | Hourglass-Decoder | [Download](https://drive.google.com/file/d/1KVINiBkVpJylx_6z1lAC7CQ4kmn-RJRN/view?usp=drive_link) |
+| v2-S | DINO2reg-ViT-Small  |    RAFT-4iter     | [Download](https://drive.google.com/file/d/1YfmvXwpWmhLg3jSxnhT7LvY0yawlXcr_/view?usp=drive_link) |
+| v2-L | DINO2reg-ViT-Large  |    RAFT-8iter     | [Download](https://drive.google.com/file/d/1eT2gG-kwsVzNy5nJrbm4KC-9DbNKyLnr/view?usp=drive_link) |
+| v2-g | DINO2reg-ViT-giant2 |    RAFT-8iter     | [Download 🤗](https://huggingface.co/JUGGHM/Metric3D/blob/main/metric_depth_vit_giant2_800k.pth) |
+
+### Dataset Mode
+1. put the trained ckpt file ```model.pth``` in ```weight/```.
+2. generate data annotation by following the code ```data/gene_annos_kitti_demo.py```, which includes 'rgb', (optional) 'intrinsic', (optional) 'depth', (optional) 'depth_scale'.
+3. change the 'test_data_path' in ```test_*.sh``` to the ```*.json``` path. 
+4. run ```source test_kitti.sh``` or ```source test_nyu.sh```.
+
+### In-the-Wild Mode
+1. put the trained ckpt file ```model.pth``` in ```weight/```.
+2. change the 'test_data_path' in ```test.sh``` to the image folder path. 
+3. run ```source test_vit.sh``` for transformers and ```source test.sh``` for convnets.
+As no intrinsics are provided, we provided by default 9 settings of focal length.
+
+### Metric3D and Droid-Slam
+If you are interested in combining metric3D and monocular visual slam system to achieve the metric slam, you can refer to this [repo](https://github.com/Jianxff/droid_metric).
+
+## ❓ Q & A
+### Q1: Why depth maps look good but pointclouds are distorted?
+Because the focal length is not properly set! Please find a proper focal length by modifying codes [here](mono/utils/do_test.py#309) yourself.  
+
+### Q2: Why the point clouds are too slow to be generated?
+Because the images are too large! Use smaller ones instead. 
+
+### Q3: Why predicted depth maps are not satisfactory?
+First be sure all black padding regions at image boundaries are cropped out. Then please try again.
+Besides, metric 3D is not almighty. Some objects (chandeliers, drones...) / camera views (aerial view, bev...) do not occur frequently in the training datasets. We will going deeper into this and release more powerful solutions.
+
+## 📧 Citation
+If you use this toolbox in your research or wish to refer to the baseline results published here, please use the following BibTeX entries:
+```
+@misc{Metric3D,
+  author =       {Yin, Wei and Hu, Mu},
+  title =        {OpenMetric3D: An Open Toolbox for Monocular Depth Estimation},
+  howpublished = {\url{https://github.com/YvanYin/Metric3D}},
+  year =         {2024}
+}
+```
+<!-- ```
+@article{hu2024metric3dv2,
+  title={Metric3D v2: A Versatile Monocular Geometric Foundation Model for Zero-shot Metric Depth and Surface Normal Estimation},
+  author={Hu, Mu and Yin, Wei and Zhang, Chi and Cai, Zhipeng and Long, Xiaoxiao and Chen, Hao and Wang, Kaixuan and Yu, Gang and Shen, Chunhua and Shen, Shaojie},
+  journal={arXiv preprint arXiv:2404.15506},
+  year={2024}
+}
+``` -->
+Also please cite our papers if this help your research.
+```
+@article{hu2024metric3dv2,
+  title={Metric3d v2: A versatile monocular geometric foundation model for zero-shot metric depth and surface normal estimation},
+  author={Hu, Mu and Yin, Wei and Zhang, Chi and Cai, Zhipeng and Long, Xiaoxiao and Chen, Hao and Wang, Kaixuan and Yu, Gang and Shen, Chunhua and Shen, Shaojie},
+  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
+  year={2024},
+  publisher={IEEE}
+}
+```
+```
+@article{yin2023metric,
+  title={Metric3D: Towards Zero-shot Metric 3D Prediction from A Single Image},
+  author={Wei Yin, Chi Zhang, Hao Chen, Zhipeng Cai, Gang Yu, Kaixuan Wang, Xiaozhi Chen, Chunhua Shen},
+  booktitle={ICCV},
+  year={2023}
+}
+```
+
+## License and Contact
+
+The *Metric 3D* code is under a 2-clause BSD License. For further commercial inquiries, please contact Dr. Wei Yin  [yvanwy@outlook.com] and Mr. Mu Hu [mhuam@connect.ust.hk].

external/Metric3D/hubconf.py

@@ -0,0 +1,227 @@
+dependencies = ['torch', 'torchvision']
+
+import os
+import torch
+try:
+  from mmcv.utils import Config, DictAction
+except:
+  from mmengine import Config, DictAction
+
+from mono.model.monodepth_model import get_configured_monodepth_model
+metric3d_dir = os.path.dirname(__file__)
+
+MODEL_TYPE = {
+  'ConvNeXt-Tiny': {
+    'cfg_file': f'{metric3d_dir}/mono/configs/HourglassDecoder/convtiny.0.3_150.py',
+    'ckpt_file': 'https://huggingface.co/JUGGHM/Metric3D/resolve/main/convtiny_hourglass_v1.pth',
+  },
+  'ConvNeXt-Large': {
+    'cfg_file': f'{metric3d_dir}/mono/configs/HourglassDecoder/convlarge.0.3_150.py',
+    'ckpt_file': 'https://huggingface.co/JUGGHM/Metric3D/resolve/main/convlarge_hourglass_0.3_150_step750k_v1.1.pth',
+  },
+  'ViT-Small': {
+    'cfg_file': f'{metric3d_dir}/mono/configs/HourglassDecoder/vit.raft5.small.py',
+    'ckpt_file': 'https://huggingface.co/JUGGHM/Metric3D/resolve/main/metric_depth_vit_small_800k.pth',
+  },
+  'ViT-Large': {
+    'cfg_file': f'{metric3d_dir}/mono/configs/HourglassDecoder/vit.raft5.large.py',
+    'ckpt_file': 'https://huggingface.co/JUGGHM/Metric3D/resolve/main/metric_depth_vit_large_800k.pth',
+  },
+  'ViT-giant2': {
+    'cfg_file': f'{metric3d_dir}/mono/configs/HourglassDecoder/vit.raft5.giant2.py',
+    'ckpt_file': 'https://huggingface.co/JUGGHM/Metric3D/resolve/main/metric_depth_vit_giant2_800k.pth',
+  },
+}
+
+
+
+def metric3d_convnext_tiny(pretrain=False, **kwargs):
+  '''
+  Return a Metric3D model with ConvNeXt-Large backbone and Hourglass-Decoder head.
+  For usage examples, refer to: https://github.com/YvanYin/Metric3D/blob/main/hubconf.py
+  Args:
+    pretrain (bool): whether to load pretrained weights.
+  Returns:
+    model (nn.Module): a Metric3D model.
+  '''
+  cfg_file = MODEL_TYPE['ConvNeXt-Tiny']['cfg_file']
+  ckpt_file = MODEL_TYPE['ConvNeXt-Tiny']['ckpt_file']
+
+  cfg = Config.fromfile(cfg_file)
+  model = get_configured_monodepth_model(cfg)
+  if pretrain:
+    model.load_state_dict(
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      strict=False,
+    )
+  return model
+
+def metric3d_convnext_large(pretrain=False, **kwargs):
+  '''
+  Return a Metric3D model with ConvNeXt-Large backbone and Hourglass-Decoder head.
+  For usage examples, refer to: https://github.com/YvanYin/Metric3D/blob/main/hubconf.py
+  Args:
+    pretrain (bool): whether to load pretrained weights.
+  Returns:
+    model (nn.Module): a Metric3D model.
+  '''
+  cfg_file = MODEL_TYPE['ConvNeXt-Large']['cfg_file']
+  ckpt_file = MODEL_TYPE['ConvNeXt-Large']['ckpt_file']
+
+  cfg = Config.fromfile(cfg_file)
+  model = get_configured_monodepth_model(cfg)
+  if pretrain:
+    model.load_state_dict(
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      strict=False,
+    )
+  return model
+
+def metric3d_vit_small(pretrain=False, **kwargs):
+  '''
+  Return a Metric3D model with ViT-Small backbone and RAFT-4iter head.
+  For usage examples, refer to: https://github.com/YvanYin/Metric3D/blob/main/hubconf.py
+  Args:
+    pretrain (bool): whether to load pretrained weights.
+  Returns:
+    model (nn.Module): a Metric3D model.
+  '''
+  cfg_file = MODEL_TYPE['ViT-Small']['cfg_file']
+  ckpt_file = MODEL_TYPE['ViT-Small']['ckpt_file']
+
+  cfg = Config.fromfile(cfg_file)
+  model = get_configured_monodepth_model(cfg)
+  if pretrain:
+    model.load_state_dict(
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      strict=False,
+    )
+  return model
+
+def metric3d_vit_large(pretrain=False, **kwargs):
+  '''
+  Return a Metric3D model with ViT-Large backbone and RAFT-8iter head.
+  For usage examples, refer to: https://github.com/YvanYin/Metric3D/blob/main/hubconf.py
+  Args:
+    pretrain (bool): whether to load pretrained weights.
+  Returns:
+    model (nn.Module): a Metric3D model.
+  '''
+  cfg_file = MODEL_TYPE['ViT-Large']['cfg_file']
+  ckpt_file = MODEL_TYPE['ViT-Large']['ckpt_file']
+
+  cfg = Config.fromfile(cfg_file)
+  model = get_configured_monodepth_model(cfg)
+  if pretrain:
+    model.load_state_dict(
+      torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+      strict=False,
+    )
+  return model
+
+def metric3d_vit_giant2(pretrain=False, **kwargs):
+  '''
+  Return a Metric3D model with ViT-Giant2 backbone and RAFT-8iter head.
+  For usage examples, refer to: https://github.com/YvanYin/Metric3D/blob/main/hubconf.py
+  Args:
+    pretrain (bool): whether to load pretrained weights.
+  Returns:
+    model (nn.Module): a Metric3D model.
+  '''
+  cfg_file = MODEL_TYPE['ViT-giant2']['cfg_file']
+  ckpt_file = MODEL_TYPE['ViT-giant2']['ckpt_file']
+  print("ckpt_file", ckpt_file)
+  cfg = Config.fromfile(cfg_file)
+  model = get_configured_monodepth_model(cfg)
+  if pretrain:
+    # model.load_state_dict(
+    #   torch.hub.load_state_dict_from_url(ckpt_file)['model_state_dict'], 
+    #   strict=False,
+    # )
+    ckpt_path = "/mnt/prev_nas/qhy_1/GenSpace/osdsynth/external/Metric3D/weight/metric_depth_vit_giant2_800k.pth"
+    state_dict = torch.load(ckpt_path, map_location="cpu")['model_state_dict']
+    model.load_state_dict(state_dict, strict=False,)
+  return model
+
+
+
+if __name__ == '__main__':
+  import cv2
+  import numpy as np
+  #### prepare data
+  rgb_file = 'data/kitti_demo/rgb/0000000050.png'
+  depth_file = 'data/kitti_demo/depth/0000000050.png'
+  intrinsic = [707.0493, 707.0493, 604.0814, 180.5066]
+  gt_depth_scale = 256.0
+  rgb_origin = cv2.imread(rgb_file)[:, :, ::-1]
+
+  #### ajust input size to fit pretrained model
+  # keep ratio resize
+  input_size = (616, 1064) # for vit model
+  # input_size = (544, 1216) # for convnext model
+  h, w = rgb_origin.shape[:2]
+  scale = min(input_size[0] / h, input_size[1] / w)
+  rgb = cv2.resize(rgb_origin, (int(w * scale), int(h * scale)), interpolation=cv2.INTER_LINEAR)
+  # remember to scale intrinsic, hold depth
+  intrinsic = [intrinsic[0] * scale, intrinsic[1] * scale, intrinsic[2] * scale, intrinsic[3] * scale]
+  # padding to input_size
+  padding = [123.675, 116.28, 103.53]
+  h, w = rgb.shape[:2]
+  pad_h = input_size[0] - h
+  pad_w = input_size[1] - w
+  pad_h_half = pad_h // 2
+  pad_w_half = pad_w // 2
+  rgb = cv2.copyMakeBorder(rgb, pad_h_half, pad_h - pad_h_half, pad_w_half, pad_w - pad_w_half, cv2.BORDER_CONSTANT, value=padding)
+  pad_info = [pad_h_half, pad_h - pad_h_half, pad_w_half, pad_w - pad_w_half]
+
+  #### normalize
+  mean = torch.tensor([123.675, 116.28, 103.53]).float()[:, None, None]
+  std = torch.tensor([58.395, 57.12, 57.375]).float()[:, None, None]
+  rgb = torch.from_numpy(rgb.transpose((2, 0, 1))).float()
+  rgb = torch.div((rgb - mean), std)
+  rgb = rgb[None, :, :, :].cuda()
+
+  ###################### canonical camera space ######################
+  # inference
+  model = torch.hub.load('yvanyin/metric3d', 'metric3d_vit_small', pretrain=True)
+  model.cuda().eval()
+  with torch.no_grad():
+    pred_depth, confidence, output_dict = model.inference({'input': rgb})
+
+  # un pad
+  pred_depth = pred_depth.squeeze()
+  pred_depth = pred_depth[pad_info[0] : pred_depth.shape[0] - pad_info[1], pad_info[2] : pred_depth.shape[1] - pad_info[3]]
+  
+  # upsample to original size
+  pred_depth = torch.nn.functional.interpolate(pred_depth[None, None, :, :], rgb_origin.shape[:2], mode='bilinear').squeeze()
+  ###################### canonical camera space ######################
+
+  #### de-canonical transform
+  canonical_to_real_scale = intrinsic[0] / 1000.0 # 1000.0 is the focal length of canonical camera
+  pred_depth = pred_depth * canonical_to_real_scale # now the depth is metric
+  pred_depth = torch.clamp(pred_depth, 0, 300)
+
+  #### you can now do anything with the metric depth 
+  # such as evaluate predicted depth
+  if depth_file is not None:
+    gt_depth = cv2.imread(depth_file, -1)
+    gt_depth = gt_depth / gt_depth_scale
+    gt_depth = torch.from_numpy(gt_depth).float().cuda()
+    assert gt_depth.shape == pred_depth.shape
+    
+    mask = (gt_depth > 1e-8)
+    abs_rel_err = (torch.abs(pred_depth[mask] - gt_depth[mask]) / gt_depth[mask]).mean()
+    print('abs_rel_err:', abs_rel_err.item())
+
+  #### normal are also available
+  if 'prediction_normal' in output_dict: # only available for Metric3Dv2, i.e. vit model
+    pred_normal = output_dict['prediction_normal'][:, :3, :, :]
+    normal_confidence = output_dict['prediction_normal'][:, 3, :, :] # see https://arxiv.org/abs/2109.09881 for details
+    # un pad and resize to some size if needed
+    pred_normal = pred_normal.squeeze()
+    pred_normal = pred_normal[:, pad_info[0] : pred_normal.shape[1] - pad_info[1], pad_info[2] : pred_normal.shape[2] - pad_info[3]]
+    # you can now do anything with the normal
+    # such as visualize pred_normal
+    pred_normal_vis = pred_normal.cpu().numpy().transpose((1, 2, 0))
+    pred_normal_vis = (pred_normal_vis + 1) / 2
+    cv2.imwrite('normal_vis.png', (pred_normal_vis * 255).astype(np.uint8))

external/Metric3D/requirements_v1.txt

@@ -0,0 +1,15 @@
+torch
+torchvision
+opencv-python
+numpy
+Pillow
+DateTime
+matplotlib
+plyfile
+HTML4Vision
+timm
+tensorboardX
+imgaug
+iopath
+imagecorruptions
+mmcv

external/Metric3D/requirements_v2.txt

@@ -0,0 +1,16 @@
+torch == 2.0.1
+torchvision == 0.15.2
+opencv-python
+numpy == 1.23.1
+xformers == 0.0.21
+Pillow
+DateTime
+matplotlib
+plyfile
+HTML4Vision
+timm
+tensorboardX
+imgaug
+iopath
+imagecorruptions
+mmcv

external/Metric3D/test.sh

@@ -0,0 +1,15 @@
+# ConvNeXt Large
+python mono/tools/test_scale_cano.py \
+    'mono/configs/HourglassDecoder/convlarge.0.3_150.py' \
+    --load-from ./weight/convlarge_hourglass_0.3_150_step750k_v1.1.pth \
+    --test_data_path ./data/wild_demo \
+    --launcher None \
+    --batch_size 2
+
+# ConvNeXt Tiny, note: only trained on outdoor data, perform better in outdoor scenes, such as kitti
+python mono/tools/test_scale_cano.py \
+    'mono/configs/HourglassDecoder/convtiny.0.3_150.py' \
+    --load-from ./weight/convtiny_hourglass_v1.pth \
+    --test_data_path ./data/wild_demo \
+    --launcher None \
+    --batch_size 2

external/Metric3D/test_kitti.sh

@@ -0,0 +1,5 @@
+python mono/tools/test_scale_cano.py \
+    'mono/configs/HourglassDecoder/test_kitti_convlarge_hourglass_0.3_150.py' \
+    --load-from ./weight/convlarge_hourglass_0.3_150_step750k_v1.1.pth \
+    --test_data_path ./data/kitti_demo/test_annotations.json \
+    --launcher None

external/Metric3D/test_nyu.sh

@@ -0,0 +1,5 @@
+python mono/tools/test_scale_cano.py \
+    'mono/configs/HourglassDecoder/test_nyu_convlarge.0.3_150.py' \
+    --load-from ./weight/convlarge_hourglass_0.3_150_step750k_v1.1.pth \
+    --test_data_path ./data/nyu_demo/test_annotations.json \
+    --launcher None

external/Metric3D/test_vit.sh

@@ -0,0 +1,5 @@
+python mono/tools/test_scale_cano.py \
+    'mono/configs/HourglassDecoder/vit.raft5.small.py' \
+    --load-from ./weight/metric_depth_vit_small_800k.pth \
+    --test_data_path ./data/wild_demo \
+    --launcher None

external/WildCamera/LICENSE

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

external/WildCamera/README.md

@@ -0,0 +1,310 @@
+## WildCamera
+This repository contains the code for the paper: **[Tame a Wild Camera: In-the-Wild Monocular Camera Calibration](https://arxiv.org/abs/2306.10988)** in NeurIPS 2023.
+<br>
+Authors: [Shengjie Zhu](https://shngjz.github.io/), [Abhinav Kumar](https://sites.google.com/view/abhinavkumar), [Masa Hu](https://scholar.google.com/citations?user=Xs-NkFMAAAAJ&hl=en), and [Xiaoming Liu](https://www.cse.msu.edu/~liuxm/index2.html)
+<br>
+[[arXiv preprint]](https://arxiv.org/abs/2306.10988)  [[Prject Page]](https://shngjz.github.io/WildCamera.github.io/) [[Poster]](https://drive.google.com/file/d/1y8v0jBd6MFtP8urHNBCzh0wsK43djIj0/view?usp=sharing)
+
+## Applications and Qualitative Results
+- 4 DoF Camera Calibration (Zero-Shot)
+  <details>
+
+  -  Camera Calibration:
+
+    https://github.com/ShngJZ/WildCamera/assets/128062217/748cf660-aebd-4a86-8d94-2be28650853b
+
+  -  DollyZoom-Demo1:
+    
+    https://github.com/ShngJZ/WildCamera/assets/128062217/15b18902-9c18-460d-8b5e-7d728cbd63c0
+
+
+  -  DollyZoom-Demo2:
+
+    https://github.com/ShngJZ/WildCamera/assets/128062217/5722039d-d0c0-49db-a7a1-c83c5e69f7fd
+
+  -  DollyZoom-Demo3:
+    
+    https://github.com/ShngJZ/WildCamera/assets/128062217/ef352b58-3e30-4b00-add8-6db5ae1d5de0
+
+- Image Crop and Resize Detection and Restoration (Zero-Shot)
+  <details>
+
+  https://github.com/ShngJZ/WildCamera/assets/128062217/c390588f-63e2-4611-b546-b86946f3caf9
+  
+- In-the-Wild Monocular 3D Object Detection ([Omni3d](https://github.com/facebookresearch/omni3d))
+  <details>
+
+  https://github.com/ShngJZ/WildCamera/assets/128062217/d776e3d0-11c3-48c2-9a1b-e5adc10408ba
+
+## Brief Introduction
+<img src="asset/framework.png" width="1000" >
+We calibrate 4 DoF intrinsic for in-the-wild images.
+The work systematically presents the connection between intrinsic and monocular 3D priors, e.g. intrinsic is inferrable from monocular depth and surface normals.
+We additionally introduce an alternative monocular 3D prior, the incidence field, for calibration.
+
+## Data Preparation
+Pretrained models and data are held in [Hugging Face](https://huggingface.co/datasets/Shengjie/WildCamera/tree/main).
+```
+WildCamera
+├── model_zoo
+│   ├── Release
+│   │   ├── wild_camera_all.pth
+│   │   ├── wild_camera_gsv.pth
+│   ├── swin_transformer
+│   │   ├── swin_large_patch4_window7_224_22k.pth
+├── data
+│   ├── MonoCalib
+│   │   ├── ARKitScenes
+│   │   ├── BIWIRGBDID
+│   │   ├── CAD120
+│   │   ├── ...
+│   │   ├── Waymo
+│   ├── UncalibTwoViewPoseEvaluation
+│   │   ├── megadepth_test_1500
+│   │   ├── scannet_test_1500
+```
+Use the script to download data in your preferred location. 
+Entire dataset takes around 150 GB disk space.
+```bash
+mkdir model_zoo
+./asset/download_wildcamera_checkpoint.sh
+ln -s your-data-location data
+./asset/download_wildcamera_dataset.sh
+```
+
+## Installation
+```bash
+conda create -n wildacamera
+conda activate wildacamera
+conda install pytorch==1.13.0 torchvision==0.14.0 torchaudio==0.13.0 pytorch-cuda=11.6 -c pytorch -c nvidia
+pip install mmcv==2.0.0 -f https://download.openmmlab.com/mmcv/dist/cu116/torch1.13/index.html
+pip install timm tensorboard loguru einops natsort h5py tabulate
+```
+You can choose difference pytorch and cuda version, however, need to follow this [link](https://mmcv.readthedocs.io/en/latest/get_started/installation.html) in selecting corresponded mmcv version.
+
+
+
+## Demo
+``` bash
+# Download demo images
+sh asset/download_demo_images.sh
+
+# Estimate intrinsic over images collected from github
+python demo/demo_inference.py
+
+# Demo inference on dolly zoom videos
+python demo/demo_dollyzoom.py
+
+# Demo image restoration
+python demo/demo_restoration.py
+```
+
+## Usage
+1. Use torch.hub to load the model (in-the-wild experiment checkpoint):
+``` bash
+model = torch.hub.load('ShngJZ/WildCamera', "WildCamera", pretrained=True)
+```
+2. Calibrate intrinsic and restore image (if needed):
+``` bash
+import PIL.Image as Image
+rgb = Image.open(path-to-image)
+
+# 4 DoF intrinsic
+intrinsic, _ = model.inference(rgb, wtassumption=False)
+# 1 DoF intrinsic
+intrinsic, _ = model.inference(rgb, wtassumption=True)
+
+# If need to restore image
+rgb_restored = model.restore_image(rgb, intrinsic, fixcrop=True)
+```
+
+## Benchmark
+``` bash
+# Benchmark Tab.2 and Tab.4
+python WildCamera/benchmark/benchmark_calibration.py --experiment_name in_the_wild
+
+# Benchmark Tab.3
+python WildCamera/benchmark/benchmark_calibration.py --experiment_name gsv
+
+# Benchmark Tab.5
+python WildCamera/benchmark/benchmark_crop.py
+
+# Benchmark Tab.6
+python WildCamera/benchmark/benchmark_uncalibtwoview_megadepth.py
+python WildCamera/benchmark/benchmark_uncalibtwoview_scannet.py
+```
+## Training
+``` bash
+# In-the-Wild Experiment
+CUDA_VISIBLE_DEVICES=0,1 python WildCamera/train/train_calibrator.py \
+--experiment_name calbr_in_the_wild \
+--experiment_set in_the_wild \
+--steps_per_epoch 2500
+
+# GSV Experiment
+CUDA_VISIBLE_DEVICES=0,1 python WildCamera/train/train_calibrator.py \
+--experiment_name calbr_gsv \
+--experiment_set gsv
+```
+
+## Citation <a name="citing"></a>
+
+Please use the following BibTeX to cite our work.
+
+```BibTeX
+@inproceedings{zhu2023tame,
+  author =       {Shengjie Zhu and Abhinav Kumar and Masa Hu and Xiaoming Liu},
+  title =        {Tame a Wild Camera: In-the-Wild Monocular Camera Calibration},
+  booktitle =    {NeurIPS},
+  year =         {2023},
+}
+```
+
+
+If you use the Tame-a-Wild-Camera benchmark, we kindly ask you to additionally cite all datasets. BibTex entries are provided below.
+
+<details><summary>Dataset BibTex</summary>
+
+```BibTex
+@inproceedings{
+  dehghan2021arkitscenes,
+  title={{ARK}itScenes - A Diverse Real-World Dataset for 3D Indoor Scene Understanding Using Mobile {RGB}-D Data},
+  author={Gilad Baruch and Zhuoyuan Chen and Afshin Dehghan and Tal Dimry and Yuri Feigin and Peter Fu and Thomas Gebauer and Brandon Joffe and Daniel Kurz and Arik Schwartz and Elad Shulman},
+  booktitle={Thirty-fifth Conference on Neural Information Processing Systems Datasets and Benchmarks Track (Round 1)},
+  year={2021},
+  url={https://openreview.net/forum?id=tjZjv_qh_CE}
+}
+```
+```BibTex
+@inproceedings{cordts2016cityscapes,
+  title={The cityscapes dataset for semantic urban scene understanding},
+  author={Cordts, Marius and Omran, Mohamed and Ramos, Sebastian and Rehfeld, Timo and Enzweiler, Markus and Benenson, Rodrigo and Franke, Uwe and Roth, Stefan and Schiele, Bernt},
+  booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
+  pages={3213--3223},
+  year={2016}
+}
+``` 
+```BibTex
+@inproceedings{geiger2012we,
+  title={Are we ready for autonomous driving? the kitti vision benchmark suite},
+  author={Geiger, Andreas and Lenz, Philip and Urtasun, Raquel},
+  booktitle={2012 IEEE conference on computer vision and pattern recognition},
+  pages={3354--3361},
+  year={2012},
+  organization={IEEE}
+}
+``` 
+```BibTex
+@inproceedings{li2018megadepth,
+  title={Megadepth: Learning single-view depth prediction from internet photos},
+  author={Li, Zhengqi and Snavely, Noah},
+  booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
+  pages={2041--2050},
+  year={2018}
+}
+``` 
+```BibTex
+@inproceedings{yu2023mvimgnet,
+  title={Mvimgnet: A large-scale dataset of multi-view images},
+  author={Yu, Xianggang and Xu, Mutian and Zhang, Yidan and Liu, Haolin and Ye, Chongjie and Wu, Yushuang and Yan, Zizheng and Zhu, Chenming and Xiong, Zhangyang and Liang, Tianyou and others},
+  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
+  pages={9150--9161},
+  year={2023}
+}
+``` 
+```BibTex
+@article{fuhrmann2014mve,
+  title={Mve-a multi-view reconstruction environment.},
+  author={Fuhrmann, Simon and Langguth, Fabian and Goesele, Michael},
+  journal={GCH},
+  volume={3},
+  pages={4},
+  year={2014}
+}
+``` 
+```BibTex
+@inproceedings{caesar2020nuscenes,
+  title={nuscenes: A multimodal dataset for autonomous driving},
+  author={Caesar, Holger and Bankiti, Varun and Lang, Alex H and Vora, Sourabh and Liong, Venice Erin and Xu, Qiang and Krishnan, Anush and Pan, Yu and Baldan, Giancarlo and Beijbom, Oscar},
+  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
+  pages={11621--11631},
+  year={2020}
+}
+``` 
+```BibTex
+@inproceedings{Silberman:ECCV12,
+  author    = {Nathan Silberman, Derek Hoiem, Pushmeet Kohli and Rob Fergus},
+  title     = {Indoor Segmentation and Support Inference from RGBD Images},
+  booktitle = {ECCV},
+  year      = {2012}
+}
+``` 
+```BibTex
+@inproceedings{ahmadyan2021objectron,
+  title={Objectron: A large scale dataset of object-centric videos in the wild with pose annotations},
+  author={Ahmadyan, Adel and Zhang, Liangkai and Ablavatski, Artsiom and Wei, Jianing and Grundmann, Matthias},
+  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
+  pages={7822--7831},
+  year={2021}
+}
+``` 
+```BibTex
+@inproceedings{sturm2012benchmark,
+  title={A benchmark for the evaluation of RGB-D SLAM systems},
+  author={Sturm, J{\"u}rgen and Engelhard, Nikolas and Endres, Felix and Burgard, Wolfram and Cremers, Daniel},
+  booktitle={2012 IEEE/RSJ international conference on intelligent robots and systems},
+  pages={573--580},
+  year={2012},
+  organization={IEEE}
+}
+``` 
+```BibTex
+@inproceedings{dai2017scannet,
+  title={Scannet: Richly-annotated 3d reconstructions of indoor scenes},
+  author={Dai, Angela and Chang, Angel X and Savva, Manolis and Halber, Maciej and Funkhouser, Thomas and Nie{\ss}ner, Matthias},
+  booktitle={Proceedings of the IEEE conference on computer vision and pattern recognition},
+  pages={5828--5839},
+  year={2017}
+}
+``` 
+```BibTex
+@article{chang2015shapenet,
+  title={Shapenet: An information-rich 3d model repository},
+  author={Chang, Angel X and Funkhouser, Thomas and Guibas, Leonidas and Hanrahan, Pat and Huang, Qixing and Li, Zimo and Savarese, Silvio and Savva, Manolis and Song, Shuran and Su, Hao and others},
+  journal={arXiv preprint arXiv:1512.03012},
+  year={2015}
+}
+``` 
+```BibTex
+@inproceedings{xiao2013sun3d,
+  title={Sun3d: A database of big spaces reconstructed using sfm and object labels},
+  author={Xiao, Jianxiong and Owens, Andrew and Torralba, Antonio},
+  booktitle={Proceedings of the IEEE international conference on computer vision},
+  pages={1625--1632},
+  year={2013}
+}
+``` 
+```BibTex
+@inproceedings{sun2020scalability,
+  title={Scalability in perception for autonomous driving: Waymo open dataset},
+  author={Sun, Pei and Kretzschmar, Henrik and Dotiwalla, Xerxes and Chouard, Aurelien and Patnaik, Vijaysai and Tsui, Paul and Guo, James and Zhou, Yin and Chai, Yuning and Caine, Benjamin and others},
+  booktitle={Proceedings of the IEEE/CVF conference on computer vision and pattern recognition},
+  pages={2446--2454},
+  year={2020}
+}
+``` 
+</details>
+
+
+
+
+
+
+
+
+
+
+
+
+
+

external/WildCamera/hubconf.py

@@ -0,0 +1,18 @@
+dependencies = ['torch']
+import torch
+from WildCamera.newcrfs.newcrf_incidencefield import NEWCRFIF
+
+def WildCamera(pretrained=True):
+    model = NEWCRFIF(version='large07', pretrained=None)
+    if pretrained:
+        # pretrained_resource = "https://huggingface.co/datasets/Shengjie/WildCamera/resolve/main/checkpoint/wild_camera_all.pth"
+        # state_dict = torch.hub.load_state_dict_from_url(pretrained_resource, map_location='cpu')
+        state_dict = torch.load(
+            "/mnt/prev_nas/qhy/WildCamera/checkpoint/wild_camera_all.pth",
+            map_location="cpu"
+        )
+        model.load_state_dict(state_dict, strict=True)
+    return model
+
+if __name__ == "__main__":
+    model = torch.hub.load('ShngJZ/WildCamera', "WildCamera", pretrained=True)

processor/captions.py

@@ -0,0 +1,92 @@
+import random
+
+from osdsynth.utils.logger import SkipImageException
+
+global_qs_list = [
+    "Can you provide a detailed description of this image in one paragraph of 30 words or less?",
+    "Could you give a concise, detailed account of what's depicted in this image, all in one paragraph, aiming for no more than 30 words?",
+    "Please offer a detailed portrayal of this image, condensed into one paragraph, keeping it under 30 words.",
+    "Could you sketch out a detailed narrative of this image within a 30-word limit, all in a single paragraph?",
+    "Would you be able to distill the essence of this image into one detailed paragraph of exactly 30 words?",
+    "Can you unpack this image's details in a succinct paragraph, ensuring it's contained to up to 30 words?",
+    "Could you elaborate on what this image shows, using no more than 30 words, all within one paragraph?",
+    "In 30 words or fewer, can you dissect the details of this image, presented in a single paragraph?",
+    "Could you render a vivid description of this image within the confines of 30 words, all in one paragraph?",
+    "Please distill the details of this image into a brief yet rich description, not exceeding 30 words, all in one paragraph.",
+    "Can you encapsulate this image's details in a comprehensive paragraph, without exceeding 30 words?",
+    "Would you mind providing a detailed explanation of this image, adhering to a 30-word limit, all in one paragraph?",
+    "Could you convey the intricate details of this image in a brief composition of no more than 30 words, contained in one paragraph?",
+    "Please craft a detailed depiction of this image, ensuring it's concise with a maximum of 30 words, all within a single paragraph.",
+    "Can you delineate the specifics of this image in a succinct narrative, capped at 30 words, all in one paragraph?",
+]
+
+landmark_prompt = [
+    "In the context of: {global_caption} Try to categorize the following image into one of these categories ['indoor', 'outdoor'], use 'others' if it is not a natural image."
+]
+
+
+class CaptionImage:
+    def __init__(self, cfg, logger, device, init_lava=False):
+        self.cfg = cfg
+        self.logger = logger
+        self.device = device
+
+        # Initialize LLava, deprecated, only use placeholders
+        if init_lava:
+            from osdsynth.processor.wrappers.llava import LLavaWrapper
+
+            self.llava_processor = LLavaWrapper(cfg, logger, device)
+        else:
+            self.llava_processor = None
+
+    def process_landmark(self, image_bgr):
+        image_tensor, image_size = self.llava_processor.process_image(image_bgr)
+
+        global_qs = random.choice(global_qs_list)
+        global_caption = self.llava_processor.process_vqa(image_tensor, image_size, global_qs, 1024)
+
+        landmark_qs = random.choice(landmark_prompt)
+        landmark_qs = landmark_qs.format(global_caption=global_caption)
+        landmark_caption = self.llava_processor.process_vqa(image_tensor, image_size, landmark_qs, 50)
+
+        if "indoor" in landmark_caption.lower():
+            landmark = "indoor"
+        elif "outdoor" in landmark_caption.lower():
+            landmark = "outdoor"
+        else:
+            raise SkipImageException("LLava failed to predict the landmark.")
+
+        return landmark, global_caption
+
+    def process_local_caption(self, detections, global_caption="", use_placeholder=True, is_one=False, is_three=False):
+        n_objects = len(detections)
+        if not is_three:
+            if n_objects < 2 and is_one==False:
+                raise SkipImageException("Ddetected objects less than 2")
+            if n_objects == 0:
+                raise SkipImageException("No objects detected finally")
+        else:
+            if n_objects < 3:
+                raise SkipImageException("Ddetected objects less than 3")
+            if n_objects == 0:
+                raise SkipImageException("No objects detected finally")
+
+        for obj_idx in range(n_objects):
+            if use_placeholder:
+                # detections[obj_idx]["caption"] = f"<region{obj_idx}>"
+                detections[obj_idx]["caption"] = f"<region{obj_idx}: {detections[obj_idx]['class_name']}>"
+            else:
+                assert self.llava_processor is not None
+                detections[obj_idx]["caption"] = f"<region{obj_idx}>"
+
+                # prepare dense caption
+                cropped_image = detections[obj_idx]["image_crop_modified"]
+                image_tensor, image_size = self.llava_processor.process_image(cropped_image, is_pil_rgb=True)
+
+                local_qs_template = r""""Can you describe the {class_name} in this close-up within five words? Highlight its color, appearance, style. For example: 'Man in red hat walking', 'Wooden pallet with boxes'."""
+                local_qs = local_qs_template.format(
+                    global_caption=global_caption, class_name=detections[obj_idx]["class_name"]
+                )
+                local_caption = self.llava_processor.process_vqa(image_tensor, image_size, local_qs, 50)
+                detections[obj_idx]["dense_caption"] = local_caption.lower().replace(".", "").strip('"')
+        return detections

processor/pointcloud.py

@@ -0,0 +1,476 @@
+import random
+from collections import Counter
+
+import cv2
+import matplotlib
+import numpy as np
+import open3d as o3d
+import torch
+from osdsynth.processor.wrappers.metric3d_v2 import get_depth_model, inference_depth
+from osdsynth.processor.wrappers.perspective_fields import (
+    create_rotation_matrix,
+    get_perspective_fields_model,
+    run_perspective_fields_model,
+)
+from PIL import Image
+from scipy.spatial.transform import Rotation
+from wis3d import Wis3D
+
+
+class PointCloudReconstruction:
+    """Class to reconstruct point cloud from depth maps."""
+
+    def __init__(self, cfg, logger, device, init_models=True):
+        """Initialize the class."""
+        self.cfg = cfg
+        self.logger = logger
+        self.device = device
+        self.vis = self.cfg.vis
+
+        if init_models:
+            # Initialize the perspective_fields_model
+            self.perspective_fields_model = get_perspective_fields_model(cfg, device)
+
+            # Initialize the Camera Intrinsics Model
+            self.wilde_camera_model = torch.hub.load("osdsynth/external/WildCamera", "WildCamera", pretrained=True,source='local').to(device)
+
+            # Initialize the Metric3D_v2
+            self.depth_model = get_depth_model(device)
+        else:
+            self.perspective_fields_model = self.wilde_camera_model = self.depth_model = None
+
+    def process(self, filename, image_bgr, detections_list):
+        """Reconstruct point cloud from depth map."""
+
+        image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
+        image_rgb_pil = Image.fromarray(image_rgb)
+
+        # Run Perspective Fields, this returns the pitch, roll
+        (
+            vis_perspective_fields,
+            perspective_fields,
+        ) = run_perspective_fields_model(self.perspective_fields_model, image_bgr)
+
+        # Perspective Fields to Rotation Matrix
+        perspective_R = create_rotation_matrix(
+            roll=perspective_fields["roll"],
+            pitch=perspective_fields["pitch"],
+            yaw=0,
+            degrees=True,
+        )
+
+        # Infer camera intrinsics
+        intrinsic, _ = self.wilde_camera_model.inference(image_rgb_pil, wtassumption=False)
+
+        # Infer depth
+        metric_depth = inference_depth(image_rgb, intrinsic, self.depth_model)
+
+        # Depth to points
+        pts3d = depth_to_points(metric_depth[None], intrinsic=intrinsic)
+        cano_pts3d = depth_to_points(metric_depth[None], R=perspective_R.T, intrinsic=intrinsic)
+
+        # Translate the points to ground
+        cano_pts3d_flattened = cano_pts3d.reshape(-1, 3)
+        sorted_flattened_points = cano_pts3d_flattened[cano_pts3d_flattened[:, 2].argsort()]
+        fifty_percent_index = int(sorted_flattened_points.shape[0] * 0.5)
+        selected_nearest_points = sorted_flattened_points[:fifty_percent_index]
+        min_y = np.min(selected_nearest_points[:, 1])
+        cano_pts3d[:, :, 1] -= min_y
+        
+        max_x = np.max(cano_pts3d[:, :, 0])
+        min_x = np.min(cano_pts3d[:, :, 0])
+        max_z = np.max(cano_pts3d[:, :, 2])
+        min_z = np.min(cano_pts3d[:, :, 2])
+        xz_max_min = [max_x, min_x, max_z, min_z]
+
+        if self.vis:
+            wis3d = Wis3D(self.cfg.wis3d_folder, filename)
+            # wis3d.add_point_cloud(vertices=pts3d.reshape((-1, 3)), colors=image_rgb.reshape(-1, 3), name="pts3d")
+            wis3d.add_point_cloud(
+                vertices=cano_pts3d.reshape((-1, 3)), colors=image_rgb.reshape(-1, 3), name="cano_pts3d"
+            )
+
+        
+
+        ### Create object points ###
+        n_objects = len(detections_list)
+
+        for obj_idx in range(n_objects):
+            if detections_list[obj_idx]["xyxy"][0] > cano_pts3d.shape[1]:
+                detections_list[obj_idx]["xyxy"][0] = cano_pts3d.shape[1]-1
+            if detections_list[obj_idx]["xyxy"][1] > cano_pts3d.shape[0]:
+                detections_list[obj_idx]["xyxy"][1] = cano_pts3d.shape[0]-1
+            if detections_list[obj_idx]["xyxy"][2] > cano_pts3d.shape[1]:
+                detections_list[obj_idx]["xyxy"][2] = cano_pts3d.shape[1]-1
+            if detections_list[obj_idx]["xyxy"][3] > cano_pts3d.shape[0]:
+                detections_list[obj_idx]["xyxy"][3] = cano_pts3d.shape[0]-1
+                
+            bbox = detections_list[obj_idx]["xyxy"]
+            
+            detections_list[obj_idx]["left_edge"] = -cano_pts3d[int((int(bbox[1])+int(bbox[3]))/ 2)][0][0]
+            detections_list[obj_idx]["right_edge"] = -cano_pts3d[int((int(bbox[1])+int(bbox[3]))/ 2)][-1][0]
+
+            detections_list[obj_idx]["square"] = np.abs((int(bbox[1])-int(bbox[3]))*(int(bbox[0])-int(bbox[2])))
+            
+            xc = int(intrinsic[0, 2])
+            yc = int(intrinsic[1, 2])
+            detections_list[obj_idx]["f"] = cano_pts3d[yc][xc] * 1000
+            
+            mask = detections_list[obj_idx]["subtracted_mask"]
+            class_name = detections_list[obj_idx]["class_name"]
+
+            object_pcd = create_object_pcd(cano_pts3d, image_rgb, mask)
+
+            # The object should at least contains 5 points
+            if len(object_pcd.points) < max(self.cfg.min_points_threshold, 5):
+                print("camera_object_pcd points less than threshold, skip this detection")
+                continue
+
+            object_pcd = process_pcd(self.cfg, object_pcd)
+
+            if len(object_pcd.points) < self.cfg.min_points_threshold_after_denoise:
+                print(f"{class_name} pcd_bbox too less points ({len(object_pcd.points)}), skip this detection")
+                continue
+
+            axis_aligned_bbox, oriented_bbox = get_bounding_box(self.cfg, object_pcd)
+
+            if axis_aligned_bbox.volume() < 1e-6:
+                print(f"{class_name} pcd_bbox got small volume, skip this detection")
+                continue
+
+            detections_list[obj_idx]["pcd"] = object_pcd
+            detections_list[obj_idx]["axis_aligned_bbox"] = axis_aligned_bbox
+            detections_list[obj_idx]["oriented_bbox"] = oriented_bbox
+
+        # Filter detections to include only those with a 'pcd' key
+        filtered_detections = [det for det in detections_list if "pcd" in det]
+
+        instance_colored_pcds = color_by_instance([det["pcd"] for det in filtered_detections])
+        axis_aligned_bbox = [det["axis_aligned_bbox"] for det in filtered_detections]
+        oriented_bboxes = [det["oriented_bbox"] for det in filtered_detections]
+
+        if self.vis:
+            obj_id = 0
+            for obj_pcd, obj_aa_box, obj_or_box in zip(instance_colored_pcds, axis_aligned_bbox, oriented_bboxes):
+                class_name = filtered_detections[obj_id]["class_name"]
+                pcd_points = np.asarray(obj_pcd.points)
+                pcd_colors = np.asarray(obj_pcd.colors)
+
+                # Convert bbox to center, euler, extent
+                aa_center, aa_eulers, aa_extent = axis_aligned_bbox_to_center_euler_extent(
+                    obj_aa_box.get_min_bound(), obj_aa_box.get_max_bound()
+                )
+                or_center, or_eulers, or_extent = oriented_bbox_to_center_euler_extent(
+                    obj_or_box.center, obj_or_box.R, obj_or_box.extent
+                )
+
+                wis3d.add_point_cloud(vertices=pcd_points, colors=pcd_colors, name=f"{obj_id:02d}_{class_name}")
+                wis3d.add_boxes(
+                    positions=aa_center, eulers=aa_eulers, extents=aa_extent, name=f"{obj_id:02d}_{class_name}_aa_bbox"
+                )
+                # wis3d.add_boxes(positions=or_center, eulers=or_eulers, extents=or_extent, name=f"{obj_id:02d}_{class_name}_or_bbox")
+                obj_id += 1
+
+        angles = [perspective_fields["roll"],perspective_fields["pitch"]]
+        return filtered_detections, angles, xz_max_min
+
+
+def depth_to_points(depth, R=None, t=None, fov=None, intrinsic=None):
+    K = intrinsic
+    Kinv = np.linalg.inv(K)
+    if R is None:
+        R = np.eye(3)
+    if t is None:
+        t = np.zeros(3)
+
+    # M converts from your coordinate to PyTorch3D's coordinate system
+    M = np.eye(3)
+
+    height, width = depth.shape[1:3]
+
+    x = np.arange(width)
+    y = np.arange(height)
+    coord = np.stack(np.meshgrid(x, y), -1)
+    coord = np.concatenate((coord, np.ones_like(coord)[:, :, [0]]), -1)  # z=1
+    coord = coord.astype(np.float32)
+    coord = coord[None]  # bs, h, w, 3
+
+    D = depth[:, :, :, None, None]
+    pts3D_1 = D * Kinv[None, None, None, ...] @ coord[:, :, :, :, None]
+    # pts3D_1 live in your coordinate system. Convert them to Py3D's
+    pts3D_1 = M[None, None, None, ...] @ pts3D_1
+    # from reference to targe tviewpoint
+    pts3D_2 = R[None, None, None, ...] @ pts3D_1 + t[None, None, None, :, None]
+
+    # G converts from your coordinate to PyTorch3D's coordinate system
+    G = np.eye(3)
+    G[0, 0] = -1.0
+    G[1, 1] = -1.0
+
+    return pts3D_2[:, :, :, :3, 0][0] @ G.T
+
+
+def create_object_pcd(image_points, image_rgb, mask):
+    points = image_points[mask]
+    colors = image_rgb[mask] / 255.0
+
+    # Perturb the points a bit to avoid colinearity
+    points += np.random.normal(0, 4e-3, points.shape)
+
+    # Create an Open3D PointCloud object
+    pcd = o3d.geometry.PointCloud()
+    pcd.points = o3d.utility.Vector3dVector(points)
+    pcd.colors = o3d.utility.Vector3dVector(colors)
+    return pcd
+
+
+def get_bounding_box(cfg, pcd):
+    axis_aligned_bbox = pcd.get_axis_aligned_bounding_box()
+
+    try:
+        oriented_bbox = pcd.get_oriented_bounding_box(robust=True)
+    except RuntimeError as e:
+        print(f"Met {e}.")
+        oriented_bbox = None
+
+    return axis_aligned_bbox, oriented_bbox
+
+
+def points_to_pcd(points, colors=None):
+    pcd = o3d.geometry.PointCloud()
+    pcd.points = o3d.utility.Vector3dVector(points.reshape(-1, 3))
+    if colors is not None:
+        pcd.colors = o3d.utility.Vector3dVector(colors.reshape(-1, 3))
+    return pcd
+
+
+def process_pcd(cfg, pcd, run_dbscan=True):
+    scale = np.linalg.norm(np.asarray(pcd.points).std(axis=0)) * 3.0 + 1e-6
+    [pcd, _] = pcd.remove_statistical_outlier(nb_neighbors=20, std_ratio=2)
+    pcd = pcd.voxel_down_sample(voxel_size=max(0.01, scale / 40))
+
+    if cfg.dbscan_remove_noise and run_dbscan:
+        # print("Before dbscan:", len(pcd.points))
+        pcd = pcd_denoise_dbscan(pcd, eps=cfg.dbscan_eps, min_points=cfg.dbscan_min_points)  #
+        # print("After dbscan:", len(pcd.points))
+
+    return pcd
+
+
+def pcd_denoise_dbscan(pcd: o3d.geometry.PointCloud, eps=0.02, min_points=10) -> o3d.geometry.PointCloud:
+    ### Remove noise via clustering
+    pcd_clusters = pcd.cluster_dbscan(
+        eps=eps,
+        min_points=min_points,
+    )
+
+    # Convert to numpy arrays
+    obj_points = np.asarray(pcd.points)
+    obj_colors = np.asarray(pcd.colors)
+    pcd_clusters = np.array(pcd_clusters)
+
+    # Count all labels in the cluster
+    counter = Counter(pcd_clusters)
+
+    # Remove the noise label
+    if counter and (-1 in counter):
+        del counter[-1]
+
+    if counter:
+        # Find the label of the largest cluster
+        most_common_label, _ = counter.most_common(1)[0]
+
+        # Create mask for points in the largest cluster
+        largest_mask = pcd_clusters == most_common_label
+
+        # Apply mask
+        largest_cluster_points = obj_points[largest_mask]
+        largest_cluster_colors = obj_colors[largest_mask]
+
+        # If the largest cluster is too small, return the original point cloud
+        if len(largest_cluster_points) < 5:
+            return pcd
+
+        # Create a new PointCloud object
+        largest_cluster_pcd = o3d.geometry.PointCloud()
+        largest_cluster_pcd.points = o3d.utility.Vector3dVector(largest_cluster_points)
+        largest_cluster_pcd.colors = o3d.utility.Vector3dVector(largest_cluster_colors)
+
+        pcd = largest_cluster_pcd
+
+    return pcd
+
+
+def create_object_pcd(image_points, image_rgb, mask):
+    points = image_points[mask]
+    colors = image_rgb[mask] / 255.0
+
+    # Perturb the points a bit to avoid colinearity
+    points += np.random.normal(0, 4e-3, points.shape)
+
+    # Create an Open3D PointCloud object
+    pcd = o3d.geometry.PointCloud()
+    pcd.points = o3d.utility.Vector3dVector(points)
+    pcd.colors = o3d.utility.Vector3dVector(colors)
+    return pcd
+
+
+def color_by_instance(pcds):
+    cmap = matplotlib.colormaps.get_cmap("turbo")
+    instance_colors = cmap(np.linspace(0, 1, len(pcds)))
+    for i in range(len(pcds)):
+        pcd = pcds[i]
+        pcd.colors = o3d.utility.Vector3dVector(np.tile(instance_colors[i, :3], (len(pcd.points), 1)))
+    return pcds
+
+
+def oriented_bbox_to_center_euler_extent(bbox_center, box_R, bbox_extent):
+    center = np.asarray(bbox_center)
+    extent = np.asarray(bbox_extent)
+    eulers = Rotation.from_matrix(box_R.copy()).as_euler("XYZ")
+    return center, eulers, extent
+
+
+def axis_aligned_bbox_to_center_euler_extent(min_coords, max_coords):
+    # Calculate the center
+    center = tuple((min_val + max_val) / 2.0 for min_val, max_val in zip(min_coords, max_coords))
+
+    # Euler angles for an axis-aligned bounding box are always 0
+    eulers = (0, 0, 0)
+
+    # Calculate the extents
+    extent = tuple(max_val - min_val for min_val, max_val in zip(min_coords, max_coords))
+
+    return center, eulers, extent
+
+
+# Distance calculations
+def human_like_distance(distance_meters):
+    # Define the choices with units included, focusing on the 0.1 to 10 meters range
+    if distance_meters < 1:  # For distances less than 1 meter
+        choices = [
+            (
+                round(distance_meters * 100, 2),
+                "centimeters",
+                0.2,
+            ),  # Centimeters for very small distances
+            (
+                round(distance_meters * 39.3701, 2),
+                "inches",
+                0.8,
+            ),  # Inches for the majority of cases under 1 meter
+        ]
+    elif distance_meters < 3:  # For distances less than 3 meters
+        choices = [
+            (round(distance_meters, 2), "meters", 0.5),
+            (
+                round(distance_meters * 3.28084, 2),
+                "feet",
+                0.5,
+            ),  # Feet as a common unit within indoor spaces
+        ]
+    else:  # For distances from 3 up to 10 meters
+        choices = [
+            (
+                round(distance_meters, 2),
+                "meters",
+                0.7,
+            ),  # Meters for clarity and international understanding
+            (
+                round(distance_meters * 3.28084, 2),
+                "feet",
+                0.3,
+            ),  # Feet for additional context
+        ]
+
+    # Normalize probabilities and make a selection
+    total_probability = sum(prob for _, _, prob in choices)
+    cumulative_distribution = []
+    cumulative_sum = 0
+    for value, unit, probability in choices:
+        cumulative_sum += probability / total_probability  # Normalize probabilities
+        cumulative_distribution.append((cumulative_sum, value, unit))
+
+    # Randomly choose based on the cumulative distribution
+    r = random.random()
+    for cumulative_prob, value, unit in cumulative_distribution:
+        if r < cumulative_prob:
+            return f"{value} {unit}"
+
+    # Fallback to the last choice if something goes wrong
+    return f"{choices[-1][0]} {choices[-1][1]}"
+
+
+def calculate_distances_between_point_clouds(A, B):
+    dist_pcd1_to_pcd2 = np.asarray(A.compute_point_cloud_distance(B))
+    dist_pcd2_to_pcd1 = np.asarray(B.compute_point_cloud_distance(A))
+    combined_distances = np.concatenate((dist_pcd1_to_pcd2, dist_pcd2_to_pcd1))
+    avg_dist = np.mean(combined_distances)
+    return human_like_distance(avg_dist)
+
+
+def calculate_centroid(pcd):
+    """Calculate the centroid of a point cloud."""
+    points = np.asarray(pcd.points)
+    centroid = np.mean(points, axis=0)
+    return centroid
+
+
+def calculate_relative_positions(centroids):
+    """Calculate the relative positions between centroids of point clouds."""
+    num_centroids = len(centroids)
+    relative_positions_info = []
+
+    for i in range(num_centroids):
+        for j in range(i + 1, num_centroids):
+            relative_vector = centroids[j] - centroids[i]
+
+            distance = np.linalg.norm(relative_vector)
+            relative_positions_info.append(
+                {"pcd_pair": (i, j), "relative_vector": relative_vector, "distance": distance}
+            )
+
+    return relative_positions_info
+
+
+def get_bounding_box_height(pcd):
+    """
+    Compute the height of the bounding box for a given point cloud.
+
+    Parameters:
+    pcd (open3d.geometry.PointCloud): The input point cloud.
+
+    Returns:
+    float: The height of the bounding box.
+    """
+    aabb = pcd.get_axis_aligned_bounding_box()
+    return aabb.get_extent()[1]  # Assuming the Y-axis is the up-direction
+
+
+def compare_bounding_box_height(pcd_i, pcd_j):
+    """
+    Compare the bounding box heights of two point clouds.
+
+    Parameters:
+    pcd_i (open3d.geometry.PointCloud): The first point cloud.
+    pcd_j (open3d.geometry.PointCloud): The second point cloud.
+
+    Returns:
+    bool: True if the bounding box of pcd_i is taller than that of pcd_j, False otherwise.
+    """
+    height_i = get_bounding_box_height(pcd_i)
+    height_j = get_bounding_box_height(pcd_j)
+
+    return height_i > height_j
+
+def get_square(point_1, point_2):
+    """Calculate the square of the distance between two points."""
+    return (point_1[0]-point_2[0]) * (point_1[1]-point_2[1])
+
+def get_focal_length(intrinsic, c_point):
+    """Calculate the focal length from the intrinsic matrix."""
+    xc = intrinsic[0, 2]
+    yc = intrinsic[1, 2]
+    
+    
+    return f

processor/prompt.py

@@ -0,0 +1,277 @@
+import random
+from itertools import combinations
+
+import numpy as np
+from osdsynth.processor.prompt_utils import *
+from osdsynth.processor.prompt_T2Ibench import *
+from osdsynth.processor.prompt_ImageEditbench import *
+from osdsynth.processor.prompt_CR import *
+
+
+
+
+class T2IPromptGenerator:
+    def __init__(self, cfg, logger, device):
+        """Initialize the class."""
+        self.cfg = cfg
+        self.logger = logger
+        self.device = device
+        self.vis = True
+
+    def evaluate_predicates_on_pairs(self, detections, n_conv=3, spatial_choice=-1):
+        # 全部是SpatialBench的prompt
+        all_prompt_variants = [
+            camera_front_camera_center, # 0
+            camera_back_camera_center, # 1
+            camera_left_camera_center, # 2
+            camera_right_camera_center, # 3
+            
+            camera_front_object_center, # 4
+            camera_back_object_center, # 5
+            camera_left_object_center, # 6
+            camera_right_object_center, # 7
+            
+            object_side_by_side_same_direction, # 8
+            object_side_by_side_opposite_direction, # 9
+            object_face_to_face, # 10
+            object_back_to_back, # 11
+            
+            object_front, # 12
+            object_back, # 13
+            object_left, # 14
+            object_right, # 15
+            
+            camera_two_objects_closer, # 16
+            camera_two_objects_farther, # 17
+            camera_two_objects_left, # 18
+            camera_two_objects_right, # 19
+            
+            object_apart_0_5meter, # 20
+            object_apart_1meter, # 21
+            object_apart_1_5meter, # 22
+            object_apart_2meter, # 23
+            
+            camera_1meter_away, # 24
+            camera_2meter_away, # 25
+            camera_3meter_away, # 26
+            camera_4meter_away, # 27
+            
+            object_bigger_than1_2, # 28
+            object_higher_20cm, # 29
+            object_longer_50cm, # 30
+            object_wider_30cm, # 31
+            
+            side_by_side_front, # 32
+            side_by_side_left, # 33
+            side_by_side_right, # 34
+            side_by_side_back, # 35
+        ]
+        
+        if spatial_choice != -1:
+            all_prompt_variants = [all_prompt_variants[spatial_choice]]
+        else:
+            raise ValueError("spatial_choice must not be -1 for T2Ibench")
+            
+        
+        if spatial_choice not in [0,1,2,3,4,5,6,7,24,25,26,27]:
+            all_combinations = list(combinations(range(len(detections)), 2))
+            random.shuffle(all_combinations)
+            selected_combinations = all_combinations[:3]
+            object_pairs = [(detections[i], detections[j]) for i, j in selected_combinations]
+            
+
+
+            results = []
+            correct = 0
+
+            for A, B in object_pairs:
+                all_prompt_variants = [item for item in all_prompt_variants]
+                # selected_predicates_choices = random.sample(all_prompt_variants, n_conv)
+                selected_predicates_choices = random.sample(all_prompt_variants, len(all_prompt_variants))
+
+                for prompt_func in selected_predicates_choices:
+                    res = prompt_func(A, B)
+                    results.append((res, A, B, prompt_func.__name__))
+                    correct = correct + 1 if res[2] else correct
+                    score = res[3]
+
+            return results, correct, score
+        else:
+            A = detections[0]
+
+            results = []
+            correct = 0
+
+            all_prompt_variants = [item for item in all_prompt_variants]
+            selected_predicates_choices = random.sample(all_prompt_variants, len(all_prompt_variants))
+
+            for prompt_func in selected_predicates_choices:
+                res = prompt_func(A)
+                results.append((res, A, prompt_func.__name__))
+                correct = correct + 1 if res[2] else correct
+                score = res[3]
+
+            return results, correct, score
+        
+        
+class ImageEditPromptGenerator:
+    def __init__(self, cfg, logger, device):
+        """Initialize the class."""
+        self.cfg = cfg
+        self.logger = logger
+        self.device = device
+        self.vis = True
+
+    def evaluate_predicates_on_pairs(self, detections, n_conv=3, spatial_choice=-1):
+        # 全部是SpatialBench的prompt
+        all_prompt_variants = [
+            camera_to_front_camera_center, # 0
+            camera_to_left_camera_center, # 1
+            camera_to_right_camera_center, # 2
+            camera_to_back_camera_center, # 3
+            
+            camera_to_front_object_center, # 4
+            camera_to_left_object_center, # 5
+            camera_to_right_object_center, # 6
+            camera_to_back_object_center, # 7
+            
+            object_insert_side_by_side_same_orientation, # 8
+            object_insert_side_by_side_opposite_orientation, # 9
+            object_insert_face_to_face, # 10
+            object_insert_back_to_back, # 11
+            
+            object_insert_front_object_center, # 12
+            object_insert_left_object_center, # 13
+            object_insert_right_object_center, # 14
+            object_insert_behind_object_center, # 15
+            
+            object_insert_front_camera_center, # 16
+            object_insert_left_camera_center, # 17
+            object_insert_right_camera_center, # 18
+            object_insert_behind_camera_center, # 19
+            
+            objectmove_close_1meter, # 20
+            objectmove_far_1meter, # 21
+            objectmove_left_1meter, # 22
+            objectmove_right_1meter, # 23
+            
+            camera_forward_1meter, # 24
+            camera_leftward_1meter, # 25
+            camera_rightward_1meter, # 26
+            camera_backward_1meter, # 27
+            
+            object_make_12bigger, # 28
+            object_make_20cm_higher, # 29
+            object_make_50cm_longer, # 30
+            object_make_40cm_wider, # 31
+        ]
+        
+        if spatial_choice != -1:
+            all_prompt_variants = [all_prompt_variants[spatial_choice]]
+        else:
+            raise ValueError("spatial_choice must not be -1 for T2Ibench")
+            
+        
+        if spatial_choice not in [0,1,2,3,4,5,6,7,20,21,22,23,24,25,26,27,28,29,30,31]:
+            object_pairs = [(detections[0], detections[1])]
+            
+
+
+            results = []
+            correct = 0
+
+            for A, B in object_pairs:
+                all_prompt_variants = [item for item in all_prompt_variants]
+                # selected_predicates_choices = random.sample(all_prompt_variants, n_conv)
+                selected_predicates_choices = random.sample(all_prompt_variants, len(all_prompt_variants))
+
+                for prompt_func in selected_predicates_choices:
+                    res = prompt_func(A, B)
+                    results.append((res, A, B, prompt_func.__name__))
+                    correct = correct + 1 if res[2] else correct
+                    score = res[3]
+
+            return results, correct, score
+        else:
+            A = detections[0]
+
+            results = []
+            correct = 0
+
+            all_prompt_variants = [item for item in all_prompt_variants]
+            selected_predicates_choices = random.sample(all_prompt_variants, len(all_prompt_variants))
+
+            for prompt_func in selected_predicates_choices:
+                res = prompt_func(A)
+                results.append((res, A, prompt_func.__name__))
+                correct = correct + 1 if res[2] else correct
+                score = res[3]
+
+            return results, correct, score
+        
+class CRPromptGenerator:
+    def __init__(self, cfg, logger, device):
+        """Initialize the class."""
+        self.cfg = cfg
+        self.logger = logger
+        self.device = device
+        self.vis = True
+
+    def evaluate_predicates_on_pairs(self, detections, is_three=False, spatial_choice=-1):
+        # 全部是SpatialBench的prompt
+        all_prompt_variants_two = [
+            CR_two_front,
+            CR_two_back,
+            CR_two_left,
+            CR_two_right,
+        ]
+        
+        all_prompt_variants_three = [
+            CR_three_front,
+            CR_three_back,
+            CR_three_left,
+            CR_three_right,
+        ]
+
+            
+        
+
+        A = detections[0]
+        B = detections[1]
+        if is_three:
+            C = detections[2]
+            all_prompt_variants = all_prompt_variants_three
+        else:
+            C = None
+            all_prompt_variants = all_prompt_variants_two
+
+        
+
+        if spatial_choice != -1:
+            all_prompt_variants = [all_prompt_variants[spatial_choice]]
+        else:
+            raise ValueError("spatial_choice must not be -1 for T2Ibench")
+
+
+        results = []
+        correct = 0
+
+        
+        all_prompt_variants = [item for item in all_prompt_variants]
+        # selected_predicates_choices = random.sample(all_prompt_variants, n_conv)
+        selected_predicates_choices = random.sample(all_prompt_variants, len(all_prompt_variants))
+
+        prompt_func = selected_predicates_choices[0]
+        
+        if is_three:
+            res = prompt_func(A, B, C)
+            results.append((res, A, B, C, prompt_func.__name__))
+
+        else:
+            res = prompt_func(A, B)
+            results.append((res, A, B, prompt_func.__name__))
+
+        correct = correct + 1 if res[2] else correct
+        score = res[3]
+
+        return results, correct, score

processor/prompt_CR.py

@@ -0,0 +1,687 @@
+import random
+from itertools import combinations
+
+import numpy as np
+from osdsynth.processor.pointcloud import calculate_distances_between_point_clouds, human_like_distance
+# from osdsynth.processor.prompt_template import *
+from osdsynth.processor.prompt_utils import *
+# from osdsynth.processor.prompt_spatitalbench_template import *
+from osdsynth.processor.prompt import *
+
+
+def get_upper(theta):
+    for i in [-1,0,1,2]:
+        if theta < i * np.pi / 2 - np.pi / 4:
+            return i * np.pi / 2 - np.pi / 4
+    return 3 * np.pi / 2 - np.pi / 4
+
+
+
+def CR_two_front(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    A_P_B_last = A['A_P_B']
+
+    
+    theta1 = np.arctan2(A_P_B[2], A_P_B[0])
+    theta2 = np.arctan2(A_P_B_last[2], A_P_B_last[0])
+    theta1_upper = get_upper(theta1)
+    
+    if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+        theta1_lower = theta1_upper - np.pi / 2
+        position = True if theta1_upper > theta2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position = True if theta2 < -3/4*np.pi or theta2 > 3/4*np.pi else False
+        
+    
+    max_angle = 30
+
+    
+    angle_rad_A = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    direction = angle_rad_A < max_angle / 180 * np.pi
+    
+    check = position and direction
+    
+    question_template = f"Are [A] and [B] maintaining their original relative relationship when viewed from the front of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position:
+            w1 = 1
+        else:
+            if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+                w1 = 1 - np.min([np.abs(theta1_upper-theta2), np.abs(theta2-theta1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta2 = theta2 + 2 * np.pi if theta2 < -3/4*np.pi else theta2 # 转化到0~2pi
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta2), np.abs(theta2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w2 = 1
+        else:
+            w2 = 1 - np.abs(angle_rad_A - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 else w1 * w2
+
+    return question, answer, check, score
+
+def CR_two_back(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    A_P_B_last = A['A_P_B']
+
+    
+    theta1 = np.arctan2(A_P_B[2], A_P_B[0]) 
+    theta2 = np.arctan2(A_P_B_last[2], A_P_B_last[0]) 
+    theta1_upper = get_upper(theta1)
+    if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+        theta1_lower = theta1_upper - np.pi / 2
+        position = True if theta1_upper > theta2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position = True if theta2 < -3/4*np.pi or theta2 > 3/4*np.pi else False
+    
+    
+    max_angle = 30
+    
+    angle_rad_A = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,1])), -1.0, 1.0))
+    direction = angle_rad_A < max_angle / 180 * np.pi
+    
+    check = position and direction
+    
+    question_template = f"Are [A] and [B] maintaining their original relative relationship when viewed from the back of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position:
+            w1 = 1
+        else:
+            if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+                w1 = 1 - np.min([np.abs(theta1_upper-theta2), np.abs(theta2-theta1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta2 = theta2 + 2 * np.pi if theta2 < -3/4*np.pi else theta2 # 转化到0~2pi
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta2), np.abs(theta2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w2 = 1
+        else:
+            w2 = 1 - np.abs(angle_rad_A - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 else w1 * w2
+
+    return question, answer, check, score
+
+def CR_two_left(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    A_P_B_last = A['A_P_B']
+    
+    theta1 = np.arctan2(A_P_B[2], A_P_B[0]) 
+    theta2 = np.arctan2(A_P_B_last[2], A_P_B_last[0]) 
+    theta1_upper = get_upper(theta1)
+    if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+        theta1_lower = theta1_upper - np.pi / 2
+        position = True if theta1_upper > theta2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position = True if theta2 < -3/4*np.pi or theta2 > 3/4*np.pi else False
+    
+    
+    max_angle = 30
+
+    
+    angle_rad_A = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([-1,0,0])), -1.0, 1.0))
+    direction = angle_rad_A < max_angle / 180 * np.pi
+    
+    check = position and direction
+    
+    question_template = f"Are [A] and [B] maintaining their original relative relationship when viewed from the left of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position:
+            w1 = 1
+        else:
+            if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+                w1 = 1 - np.min([np.abs(theta1_upper-theta2), np.abs(theta2-theta1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta2 = theta2 + 2 * np.pi if theta2 < -3/4*np.pi else theta2 # 转化到0~2pi
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta2), np.abs(theta2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w2 = 1
+        else:
+            w2 = 1 - np.abs(angle_rad_A - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 else w1 * w2
+
+    return question, answer, check, score
+
+def CR_two_right(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    A_P_B_last = A['A_P_B']
+
+    
+    theta1 = np.arctan2(A_P_B[2], A_P_B[0]) 
+    theta2 = np.arctan2(A_P_B_last[2], A_P_B_last[0]) 
+    theta1_upper = get_upper(theta1)
+    if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+        theta1_lower = theta1_upper - np.pi / 2
+        position = True if theta1_upper > theta2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position = True if theta2 < -3/4*np.pi or theta2 > 3/4*np.pi else False
+    
+    
+    max_angle = 30
+    # angle_rad_AB = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # same_direction = angle_rad_AB < max_angle / 180 * np.pi
+    
+    angle_rad_A = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    direction = angle_rad_A < max_angle / 180 * np.pi
+    
+    check = position and direction
+    
+    question_template = f"Are [A] and [B] maintaining their original relative relationship when viewed from the right of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position:
+            w1 = 1
+        else:
+            if theta1_upper < np.pi and theta1_upper > -np.pi/2: 
+                w1 = 1 - np.min([np.abs(theta1_upper-theta2), np.abs(theta2-theta1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta2 = theta2 + 2 * np.pi if theta2 < -3/4*np.pi else theta2 # 转化到0~2pi
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta2), np.abs(theta2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w2 = 1
+        else:
+            w2 = 1 - np.abs(angle_rad_A - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 else w1 * w2
+
+    return question, answer, check, score
+
+
+def CR_three_front(A, B, C):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    C_desc, C_cloud = C["caption"], C["pcd"]
+    C_desc = C_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    C_pos = C_cloud.get_center()
+    C_pos[0] = -C_pos[0]; C_pos[1] = -C_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    C_rotation_matrix = C["rotation_matrix"]
+    
+    A_rotation_matrix = C_rotation_matrix.T @ A_rotation_matrix # 在C的坐标系下，A的旋转矩阵
+    B_rotation_matrix = C_rotation_matrix.T @ B_rotation_matrix # 在C的坐标系下，B的旋转矩阵
+    
+    C_P_A = C_rotation_matrix.T @ (A_pos - C_pos) # 在C的坐标系下，A相对于C的位置
+    C_P_B = C_rotation_matrix.T @ (B_pos - C_pos) # 在C的坐标系下，B相对于C的位置
+    C_P_A_last = C['C_P_A']
+    C_P_B_last = C['C_P_B']
+    
+    theta_CA1 = np.arctan2(C_P_A[2], C_P_A[0])
+    theta_CB1 = np.arctan2(C_P_B[2], C_P_B[0])
+    theta_CA2 = np.arctan2(C_P_A_last[2], C_P_A_last[0])
+    theta_CB2 = np.arctan2(C_P_B_last[2], C_P_B_last[0])
+    
+    
+    theta_CA1_upper = get_upper(theta_CA1)
+    theta_CA1_lower = theta_CA1_upper - np.pi / 2
+    if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2: 
+        theta1_lower = theta_CA1_upper - np.pi / 2
+        position_CA = True if theta_CA1_upper > theta_CA2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CA = True if theta_CA2 < -3/4*np.pi or theta_CA2 > 3/4*np.pi else False
+        
+    
+    theta_CB1_upper = get_upper(theta_CB1)
+    theta_CB1_lower = theta_CB1_upper - np.pi / 2
+    if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2: 
+        theta1_lower = theta_CB1_upper - np.pi / 2
+        position_CB = True if theta_CB1_upper > theta_CB2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CB = True if theta_CB2 < -3/4*np.pi or theta_CB2 > 3/4*np.pi else False
+    
+    
+    position = position_CA and position_CB
+    
+    
+    max_angle = 30
+
+    
+    angle_rad_C = np.arccos(np.clip(np.dot(C_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    direction = angle_rad_C < max_angle / 180 * np.pi
+
+
+    check = position and direction
+    question_template = f"Are [A], [B] and [C] maintaining their original relative relationship when viewed from the front of [C]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc).replace("[C]", C_desc)
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position_CA:
+            w1 = 1
+        else:
+            if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2:                
+                w1 = 1 - np.min([np.abs(theta_CA1_upper-theta_CA2), np.abs(theta_CA2-theta_CA1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta_CA2 = theta_CA2 + 2 * np.pi if theta_CA2 < -3/4*np.pi else theta_CA2
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta_CA2), np.abs(theta_CA2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+                
+        if position_CB:
+            w2 = 1
+        else:
+            if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2:                
+                w2 = 1 - np.min([np.abs(theta_CB1_upper-theta_CB2), np.abs(theta_CB2-theta_CB1_lower)]) / (np.pi / 6)
+            else: # 135~180和-180~-135的情况
+                theta_CB2 = theta_CB2 + 2 * np.pi if theta_CB2 < -3/4*np.pi else theta_CB2
+                w2 = 1 - np.min([np.abs(1.25*np.pi-theta_CB2), np.abs(theta_CB2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w3 = 1
+        else:
+            w3 = 1 - np.abs(angle_rad_C - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+        
+    return question, answer, check, score
+
+def CR_three_back(A, B, C):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    C_desc, C_cloud = C["caption"], C["pcd"]
+    C_desc = C_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    C_pos = C_cloud.get_center()
+    C_pos[0] = -C_pos[0]; C_pos[1] = -C_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    C_rotation_matrix = C["rotation_matrix"]
+    
+    A_rotation_matrix = C_rotation_matrix.T @ A_rotation_matrix # 在C的坐标系下，A的旋转矩阵
+    B_rotation_matrix = C_rotation_matrix.T @ B_rotation_matrix # 在C的坐标系下，B的旋转矩阵
+    
+    C_P_A = C_rotation_matrix.T @ (A_pos - C_pos) # 在C的坐标系下，A相对于C的位置
+    C_P_B = C_rotation_matrix.T @ (B_pos - C_pos) # 在C的坐标系下，B相对于C的位置
+    C_P_A_last = C['C_P_A']
+    C_P_B_last = C['C_P_B']
+    
+    theta_CA1 = np.arctan2(C_P_A[2], C_P_A[0])
+    theta_CB1 = np.arctan2(C_P_B[2], C_P_B[0])
+    theta_CA2 = np.arctan2(C_P_A_last[2], C_P_A_last[0])
+    theta_CB2 = np.arctan2(C_P_B_last[2], C_P_B_last[0])
+    
+    theta_CA1_upper = get_upper(theta_CA1)
+    theta_CA1_lower = theta_CA1_upper - np.pi / 2
+    if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2: 
+        theta1_lower = theta_CA1_upper - np.pi / 2
+        position_CA = True if theta_CA1_upper > theta_CA2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CA = True if theta_CA2 < -3/4*np.pi or theta_CA2 > 3/4*np.pi else False
+        
+    
+    theta_CB1_upper = get_upper(theta_CB1)
+    theta_CB1_lower = theta_CB1_upper - np.pi / 2
+    if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2: 
+        theta1_lower = theta_CB1_upper - np.pi / 2
+        position_CB = True if theta_CB1_upper > theta_CB2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CB = True if theta_CB2 < -3/4*np.pi or theta_CB2 > 3/4*np.pi else False
+    
+    
+    position = position_CA and position_CB
+    
+    
+    max_angle = 30
+    # angle_rad_AC = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # angle_rad_BC = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # same_direction_AC = angle_rad_AC < max_angle / 180 * np.pi
+    # same_direction_BC = angle_rad_BC < max_angle / 180 * np.pi
+    # same_direction = same_direction_AC and same_direction_BC
+    
+    
+    angle_rad_C = np.arccos(np.clip(np.dot(C_rotation_matrix.T[0], np.array([0,0,1])), -1.0, 1.0))
+    direction = angle_rad_C < max_angle / 180 * np.pi
+
+
+    check = position and direction
+    question_template = f"Are [A], [B] and [C] maintaining their original relative relationship when viewed from the back of [C]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc).replace("[C]", C_desc)
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position_CA:
+            w1 = 1
+        else:
+            if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2:                
+                w1 = 1 - np.min([np.abs(theta_CA1_upper-theta_CA2), np.abs(theta_CA2-theta_CA1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta_CA2 = theta_CA2 + 2 * np.pi if theta_CA2 < -3/4*np.pi else theta_CA2
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta_CA2), np.abs(theta_CA2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+                
+        if position_CB:
+            w2 = 1
+        else:
+            if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2:                
+                w2 = 1 - np.min([np.abs(theta_CB1_upper-theta_CB2), np.abs(theta_CB2-theta_CB1_lower)]) / (np.pi / 6)
+            else: # 135~180和-180~-135的情况
+                theta_CB2 = theta_CB2 + 2 * np.pi if theta_CB2 < -3/4*np.pi else theta_CB2
+                w2 = 1 - np.min([np.abs(1.25*np.pi-theta_CB2), np.abs(theta_CB2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w3 = 1
+        else:
+            w3 = 1 - np.abs(angle_rad_C - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+        
+    return question, answer, check, score
+
+def CR_three_left(A, B, C):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    C_desc, C_cloud = C["caption"], C["pcd"]
+    C_desc = C_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    C_pos = C_cloud.get_center()
+    C_pos[0] = -C_pos[0]; C_pos[1] = -C_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    C_rotation_matrix = C["rotation_matrix"]
+    
+    A_rotation_matrix = C_rotation_matrix.T @ A_rotation_matrix # 在C的坐标系下，A的旋转矩阵
+    B_rotation_matrix = C_rotation_matrix.T @ B_rotation_matrix # 在C的坐标系下，B的旋转矩阵
+    
+    C_P_A = C_rotation_matrix.T @ (A_pos - C_pos) # 在C的坐标系下，A相对于C的位置
+    C_P_B = C_rotation_matrix.T @ (B_pos - C_pos) # 在C的坐标系下，B相对于C的位置
+    C_P_A_last = C['C_P_A']
+    C_P_B_last = C['C_P_B']
+    
+    theta_CA1 = np.arctan2(C_P_A[2], C_P_A[0])
+    theta_CB1 = np.arctan2(C_P_B[2], C_P_B[0])
+    theta_CA2 = np.arctan2(C_P_A_last[2], C_P_A_last[0])
+    theta_CB2 = np.arctan2(C_P_B_last[2], C_P_B_last[0])
+    
+    theta_CA1_upper = get_upper(theta_CA1)
+    theta_CA1_lower = theta_CA1_upper - np.pi / 2
+    if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2: 
+        theta1_lower = theta_CA1_upper - np.pi / 2
+        position_CA = True if theta_CA1_upper > theta_CA2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CA = True if theta_CA2 < -3/4*np.pi or theta_CA2 > 3/4*np.pi else False
+        
+    
+    theta_CB1_upper = get_upper(theta_CB1)
+    theta_CB1_lower = theta_CB1_upper - np.pi / 2
+    if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2: 
+        theta1_lower = theta_CB1_upper - np.pi / 2
+        position_CB = True if theta_CB1_upper > theta_CB2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CB = True if theta_CB2 < -3/4*np.pi or theta_CB2 > 3/4*np.pi else False
+    
+    
+    position = position_CA and position_CB
+    
+    
+    max_angle = 30
+    # angle_rad_AC = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # angle_rad_BC = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # same_direction_AC = angle_rad_AC < max_angle / 180 * np.pi
+    # same_direction_BC = angle_rad_BC < max_angle / 180 * np.pi
+    # same_direction = same_direction_AC and same_direction_BC
+    
+    
+    angle_rad_C = np.arccos(np.clip(np.dot(C_rotation_matrix.T[0], np.array([-1,0,0])), -1.0, 1.0))
+    direction = angle_rad_C < max_angle / 180 * np.pi
+
+
+    check = position and direction
+    question_template = f"Are [A], [B] and [C] maintaining their original relative relationship when viewed from the left of [C]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc).replace("[C]", C_desc)
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position_CA:
+            w1 = 1
+        else:
+            if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2:                
+                w1 = 1 - np.min([np.abs(theta_CA1_upper-theta_CA2), np.abs(theta_CA2-theta_CA1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta_CA2 = theta_CA2 + 2 * np.pi if theta_CA2 < -3/4*np.pi else theta_CA2
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta_CA2), np.abs(theta_CA2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+                
+        if position_CB:
+            w2 = 1
+        else:
+            if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2:                
+                w2 = 1 - np.min([np.abs(theta_CB1_upper-theta_CB2), np.abs(theta_CB2-theta_CB1_lower)]) / (np.pi / 6)
+            else: # 135~180和-180~-135的情况
+                theta_CB2 = theta_CB2 + 2 * np.pi if theta_CB2 < -3/4*np.pi else theta_CB2
+                w2 = 1 - np.min([np.abs(1.25*np.pi-theta_CB2), np.abs(theta_CB2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w3 = 1
+        else:
+            w3 = 1 - np.abs(angle_rad_C - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+        
+    return question, answer, check, score
+
+def CR_three_right(A, B, C):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    C_desc, C_cloud = C["caption"], C["pcd"]
+    C_desc = C_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    C_pos = C_cloud.get_center()
+    C_pos[0] = -C_pos[0]; C_pos[1] = -C_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    C_rotation_matrix = C["rotation_matrix"]
+    
+    A_rotation_matrix = C_rotation_matrix.T @ A_rotation_matrix # 在C的坐标系下，A的旋转矩阵
+    B_rotation_matrix = C_rotation_matrix.T @ B_rotation_matrix # 在C的坐标系下，B的旋转矩阵
+    
+    C_P_A = C_rotation_matrix.T @ (A_pos - C_pos) # 在C的坐标系下，A相对于C的位置
+    C_P_B = C_rotation_matrix.T @ (B_pos - C_pos) # 在C的坐标系下，B相对于C的位置
+    C_P_A_last = C['C_P_A']
+    C_P_B_last = C['C_P_B']
+    
+    theta_CA1 = np.arctan2(C_P_A[2], C_P_A[0])
+    theta_CB1 = np.arctan2(C_P_B[2], C_P_B[0])
+    theta_CA2 = np.arctan2(C_P_A_last[2], C_P_A_last[0])
+    theta_CB2 = np.arctan2(C_P_B_last[2], C_P_B_last[0])
+    
+    theta_CA1_upper = get_upper(theta_CA1)
+    theta_CA1_lower = theta_CA1_upper - np.pi / 2
+    if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2: 
+        theta1_lower = theta_CA1_upper - np.pi / 2
+        position_CA = True if theta_CA1_upper > theta_CA2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CA = True if theta_CA2 < -3/4*np.pi or theta_CA2 > 3/4*np.pi else False
+        
+    
+    theta_CB1_upper = get_upper(theta_CB1)
+    theta_CB1_lower = theta_CB1_upper - np.pi / 2
+    if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2: 
+        theta1_lower = theta_CB1_upper - np.pi / 2
+        position_CB = True if theta_CB1_upper > theta_CB2 > theta1_lower else False
+    else: # 135~180和-180~-135的情况
+        position_CB = True if theta_CB2 < -3/4*np.pi or theta_CB2 > 3/4*np.pi else False
+    
+    
+    position = position_CA and position_CB
+    
+    
+    max_angle = 30
+    # angle_rad_AC = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # angle_rad_BC = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    # same_direction_AC = angle_rad_AC < max_angle / 180 * np.pi
+    # same_direction_BC = angle_rad_BC < max_angle / 180 * np.pi
+    # same_direction = same_direction_AC and same_direction_BC
+    
+    
+    angle_rad_C = np.arccos(np.clip(np.dot(C_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    direction = angle_rad_C < max_angle / 180 * np.pi
+
+
+    check = position and direction
+    question_template = f"Are [A], [B] and [C] maintaining their original relative relationship when viewed from the right of [C]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc).replace("[C]", C_desc)
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if position_CA:
+            w1 = 1
+        else:
+            if theta_CA1_upper < np.pi and theta_CA1_upper > -np.pi/2:                
+                w1 = 1 - np.min([np.abs(theta_CA1_upper-theta_CA2), np.abs(theta_CA2-theta_CA1_lower)]) / (np.pi / 6) # 30度的阈值
+            else: # 135~180和-180~-135的情况
+                theta_CA2 = theta_CA2 + 2 * np.pi if theta_CA2 < -3/4*np.pi else theta_CA2
+                w1 = 1 - np.min([np.abs(1.25*np.pi-theta_CA2), np.abs(theta_CA2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+                
+        if position_CB:
+            w2 = 1
+        else:
+            if theta_CB1_upper < np.pi and theta_CB1_upper > -np.pi/2:                
+                w2 = 1 - np.min([np.abs(theta_CB1_upper-theta_CB2), np.abs(theta_CB2-theta_CB1_lower)]) / (np.pi / 6)
+            else: # 135~180和-180~-135的情况
+                theta_CB2 = theta_CB2 + 2 * np.pi if theta_CB2 < -3/4*np.pi else theta_CB2
+                w2 = 1 - np.min([np.abs(1.25*np.pi-theta_CB2), np.abs(theta_CB2-0.75*np.pi)]) / (np.pi / 6) # 30度的阈值
+            
+        if direction:
+            w3 = 1
+        else:
+            w3 = 1 - np.abs(angle_rad_C - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+        
+    return question, answer, check, score

processor/prompt_ImageEditbench.py

@@ -0,0 +1,1055 @@
+import random
+from itertools import combinations
+
+import numpy as np
+from osdsynth.processor.pointcloud import calculate_distances_between_point_clouds, human_like_distance
+# from osdsynth.processor.prompt_template import *
+from osdsynth.processor.prompt_utils import *
+# from osdsynth.processor.prompt_spatitalbench_template import *
+from osdsynth.processor.prompt import *
+import torch
+
+def camera_to_front_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_front_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_front_view
+
+    question_template = f"Is the camera facing the front of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score    
+
+def camera_to_left_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_left_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_left_view
+
+    question_template = f"Is the camera facing the left of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score 
+
+def camera_to_right_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_right_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_right_view
+
+    question_template = f"Is the camera facing the right of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score 
+
+def camera_to_back_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_back_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_back_view
+
+    question_template = f"Is the camera facing the back of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def camera_to_front_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_front_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_front_view
+
+    question_template = f"Is the camera facing the front of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score    
+
+def camera_to_left_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_left_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_left_view
+
+    question_template = f"Is the camera facing the left of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score 
+
+def camera_to_right_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_right_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_right_view
+
+    question_template = f"Is the camera facing the right of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score 
+
+def camera_to_back_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_back_view = angle_rad < max_angle / 180 * np.pi
+
+    check = is_back_view
+
+    question_template = f"Is the camera facing the back of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def object_insert_side_by_side_same_orientation(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+
+    is_side_by_side = np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) > np.pi * 1 / 3
+
+    # 比较X轴的夹角
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_same_orientation = angle_rad < max_angle / 180 * np.pi
+    
+    check = is_same_orientation and is_side_by_side
+
+    print("is_same_orientation", is_same_orientation, "is_side_by_side", is_side_by_side)
+    question_template = f"Is [A] and [B] side by side, facing the same direction?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if is_same_orientation:
+            w1 = 1
+        else:
+            w1 = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        
+        if is_side_by_side:
+            w2 = 1
+        else:
+            w2 = 1 - 1*np.abs((np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) - np.pi * 1 / 3) / (np.pi/12))
+        score = 0 if w1 < 0 or w2<0 else w1*w2
+        if w1 == 1:
+            score = 0.5
+    return question, answer, check, score
+
+def object_insert_side_by_side_opposite_orientation(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+
+    is_side_by_side = np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) > np.pi * 1 / 3
+    # 比较X轴的夹角
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([-1,0,0])), -1.0, 1.0))
+    is_opposite_orientation = angle_rad < max_angle / 180 * np.pi
+
+    check = is_opposite_orientation and is_side_by_side
+
+    print("is_opposite_orientation", is_opposite_orientation, "is_side_by_side", is_side_by_side)
+    question_template = f"Is [A] and [B] side by side, facing the opposite direction?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        if is_opposite_orientation:
+            w1 = 1
+        else:
+            w1 = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        if is_side_by_side:
+            w2 = 1
+        else:
+            w2 = 1 - 1*np.abs((np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) - np.pi * 1 / 3) / (np.pi/12))
+        score = 0 if w1 < 0 or w2<0 else w1*w2
+        if w1 == 1:
+            score = 0.5
+    return question, answer, check, score
+
+def object_insert_face_to_face(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+    
+    
+    is_line =  B_P_A[0] > 0 and np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) < np.pi/3# 在一条线上，且A在B的前面
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], [-1,0,0]), -1.0, 1.0))
+    is_opposite_orientation = angle_rad < max_angle / 180 * np.pi
+    
+    check = is_opposite_orientation and is_line
+
+    print("is_opposite_orientation", is_opposite_orientation, "is_line", is_line)
+    question_template = f"Is [A] and [B] face to face?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if is_opposite_orientation:
+            w1 = 1
+        else:
+            w1 = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        if is_line:
+            w2 = 1
+        else:
+            w2 = 1 - 1*np.abs((np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) - np.pi * 1 / 3) / (np.pi/12))
+            if B_P_A[0] < 0 or np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) > np.pi/3:
+                w2 = 0
+        score = 0 if w1<0 or w2<0 else w1*w2
+        if w1 == 1:
+            score = 0.5
+    return question, answer, check, score
+
+def object_insert_back_to_back(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+
+
+    is_line =  B_P_A[0] < 0 and np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) < np.pi/3# 在一条线上，且A在B的前面
+    
+    max_angle = 30
+    angle_rad =  np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], [-1,0,0]), -1.0, 1.0))
+    is_opposite_orientation = angle_rad < max_angle / 180 * np.pi
+    
+    check = is_opposite_orientation and is_line
+
+    print("is_opposite_orientation", is_opposite_orientation, "is_line", is_line)
+    question_template = f"Is [A] and [B] back to back?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        if is_opposite_orientation:
+            w1 = 1
+        else:
+            w1 = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        if is_line:
+            w2 = 1
+        else:
+            w2 = 1 - 1*np.abs((np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) - np.pi * 1 / 3) / (np.pi/12))
+            if B_P_A[0] > 0 or np.abs(np.arctan(B_P_A[2]/ B_P_A[0])) > np.pi/3:
+                w2 = 0
+        score = 0 if w1<0 or w2<0 else w1*w2
+        if w1 == 1:
+            score = 0.5
+    return question, answer, check, score
+
+def object_insert_front_object_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    A_rotation_matrix = A["rotation_matrix"]
+    # B_rotation_matrix = B["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，A物体的位置
+
+
+    is_front =  A_P_B[0] > 0
+        
+    check = is_front
+    
+    question_template = f"Is [B] in front of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+        
+    return question, answer, check, score
+
+def object_insert_left_object_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    A_rotation_matrix = A["rotation_matrix"]
+    # B_rotation_matrix = B["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，A物体的位置
+    
+    max_angle = 30
+    A_P_B_direcetion = A_P_B / np.linalg.norm(A_P_B)
+    angle_rad = np.arccos(np.clip(np.dot(A_P_B_direcetion, np.array([0,0,-1])), -1.0, 1.0))
+    B_is_in_left_A = A_P_B[2] < 0 and angle_rad < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    is_left =  A_P_B[2] < 0 and B_is_in_left_A
+        
+    check = is_left
+    
+    question_template = f"Is [B] in the left of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 or A_P_B[2] > 0 else score
+
+    return question, answer, check, score  
+
+def object_insert_right_object_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    A_rotation_matrix = A["rotation_matrix"]
+    # B_rotation_matrix = B["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，A物体的位置
+    
+    max_angle = 30
+    A_P_B_direcetion = A_P_B / np.linalg.norm(A_P_B)
+    angle_rad = np.arccos(np.clip(np.dot(A_P_B_direcetion, np.array([0,0,1])), -1.0, 1.0))
+    B_is_in_right_A = A_P_B[2] > 0 and angle_rad < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    is_right =  A_P_B[2] > 0 and B_is_in_right_A
+        
+    check = is_right
+    
+    question_template = f"Is [B] in the right of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45))
+        score = 0 if score < 0 or A_P_B[2] < 0 else score
+
+    return question, answer, check, score  
+
+def object_insert_behind_object_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    A_rotation_matrix = A["rotation_matrix"]
+    # B_rotation_matrix = B["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，A物体的位置
+
+    is_behind =  A_P_B[0] < 0
+        
+    check = is_behind
+    
+    question_template = f"Is [B] behind [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    
+    return question, answer, check, score
+
+
+## 描述物体
+# 定性
+
+def object_insert_front_camera_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    is_front =  B_pos[2] < A_pos[2]
+        
+    check = is_front
+    
+    question_template = f"Is [B] in front of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+        
+    return question, answer, check, score
+
+def object_insert_left_camera_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+
+    is_left =  B_pos[0] < A_pos[0]
+        
+    check = is_left
+    
+    question_template = f"Is [B] in left of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+        
+    return question, answer, check, score
+
+def object_insert_right_camera_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    is_right =  B_pos[0] > A_pos[0]
+        
+    check = is_right
+    
+    question_template = f"Is [B] in right of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+        
+    return question, answer, check, score
+
+def object_insert_behind_camera_center(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+
+    is_behind =  B_pos[2] > A_pos[2]
+        
+    check = is_behind
+    
+    question_template = f"Is [B] in behind of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+        
+    return question, answer, check, score
+
+def objectmove_close_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[2] -A['last_pos'][2]
+    
+    delta = 1.0/3
+    gt_distance = -1
+    
+    check = (1+delta)*gt_distance < distance and distance < (1-delta)*gt_distance
+
+    question_template = f"Does [A] move 1 meter close to the camera?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def objectmove_far_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[2] -A['last_pos'][2]
+    
+    delta = 1.0/3
+    gt_distance = 1
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Does [A] move 1 meter far to the camera?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def objectmove_left_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[0] -A['last_pos'][0]
+    
+    delta = 1.0/3
+    gt_distance = -1
+    
+    check = (1+delta)*gt_distance < distance and distance < (1-delta)*gt_distance
+
+    question_template = f"Does [A] move 1 meter left?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def objectmove_right_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[0] -A['last_pos'][0]
+    
+    delta = 1.0/3
+    gt_distance = 1
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Does [A] move 1 meter right?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+
+def camera_forward_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[2] - A['last_pos'][2]
+    
+    delta = 1.0/3
+    gt_distance = -1
+    
+    check = (1+delta)*gt_distance < distance and distance < (1-delta)*gt_distance
+
+    question_template = f"Does camera move 1 meter forward? [A]"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def camera_leftward_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[0] - A['last_pos'][0]
+    
+    delta = 1.0/3
+    gt_distance = 1
+    
+    check = (1+delta)*gt_distance < distance and distance < (1-delta)*gt_distance
+
+    question_template = f"Does camera move 1 meter leftward? [A]"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def camera_rightward_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[0] - A['last_pos'][0]
+    
+    delta = 1.0/3
+    gt_distance = -1
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Does camera move 1 meter rightward? [A]"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def camera_backward_1meter(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    distance = A_pos[2] - A['last_pos'][2]
+    
+    delta = 1.0/3
+    gt_distance = 1
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Does camera move 1 meter backward? [A]"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+
+# 定量
+def object_make_12bigger(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    theta_A = np.arctan2(A_rotation_matrix.T[0][2], A_rotation_matrix.T[0][0])
+    A_center = A["pcd"].get_center()
+    R = A["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_A)) 
+    A["pcd"] = A["pcd"].rotate(R)
+    A_length = A["pcd"].get_axis_aligned_bounding_box().get_extent()[0]
+    A_height = A["pcd"].get_axis_aligned_bounding_box().get_extent()[1]
+    A_width = A["pcd"].get_axis_aligned_bounding_box().get_extent()[2]
+    volume = A_length * A_height * A_width
+    
+    distance = volume / A['last_volume'] - 1
+    
+    delta = 1.0/3
+    gt_distance = 0.2
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Does [A] become 1.5 times its initial dimensions.?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_make_20cm_higher(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 计算距离
+        
+    height = A["pcd"].get_axis_aligned_bounding_box().get_extent()[1]
+    last_height = A["last_height"]
+    distance = height-last_height 
+    
+    delta = 1.0/3
+    gt_distance = 0.2
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] higher 20cm than [B]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def object_make_50cm_longer(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 计算距离
+        
+    A_rotation_matrix = A["rotation_matrix"]
+    theta_A = np.arctan2(A_rotation_matrix.T[0][2], A_rotation_matrix.T[0][0])
+    A_center = A["pcd"].get_center()
+    R = A["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_A)) 
+    A["pcd"] = A["pcd"].rotate(R)
+    length = A["pcd"].get_axis_aligned_bounding_box().get_extent()[0]
+    
+    last_length = A["last_length"]
+    distance = length-last_length 
+    
+    delta = 1.0/3
+    gt_distance = 0.5
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] higher 20cm than [B]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def object_make_40cm_wider(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 计算距离
+        
+    A_rotation_matrix = A["rotation_matrix"]
+    theta_A = np.arctan2(A_rotation_matrix.T[0][2], A_rotation_matrix.T[0][0])
+    A_center = A["pcd"].get_center()
+    R = A["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_A)) 
+    A["pcd"] = A["pcd"].rotate(R)
+    width = A["pcd"].get_axis_aligned_bounding_box().get_extent()[2]
+    
+    last_width = A["last_width"]
+    distance = width-last_width 
+    
+    delta = 1.0/3
+    gt_distance = 0.4
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] higher 20cm than [B]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score

processor/prompt_T2Ibench.py

@@ -0,0 +1,1296 @@
+import random
+from itertools import combinations
+
+import numpy as np
+from osdsynth.processor.pointcloud import calculate_distances_between_point_clouds, human_like_distance
+# from osdsynth.processor.prompt_template import *
+from osdsynth.processor.prompt_utils import *
+# from osdsynth.processor.prompt_spatitalbench_template import *
+from osdsynth.processor.prompt import *
+
+def camera_front_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 15
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_front = angle_rad < max_angle / 180 * np.pi
+
+    check = is_front
+
+    question_template = f"Does the camera face the front of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(45-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_back_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 15
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_back = angle_rad < max_angle / 180 * np.pi
+
+    check = is_back
+
+    question_template = f"Does the camera face the back of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(45-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_left_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_left = angle_rad < max_angle / 180 * np.pi
+
+    check = is_left
+
+    question_template = f"Does the camera face the left of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(60-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_right_camera_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_right = angle_rad < max_angle / 180 * np.pi
+
+    check = is_right
+
+    question_template = f"Does the camera face the right of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(60-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+
+
+def camera_front_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 15
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_front = angle_rad < max_angle / 180 * np.pi
+
+    check = is_front
+
+    question_template = f"Does the camera face the front of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(45-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_back_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转���到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 15
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_back = angle_rad < max_angle / 180 * np.pi
+
+    check = is_back
+
+    question_template = f"Does the camera face the back of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(45-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_left_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(-A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_left = angle_rad < max_angle / 180 * np.pi
+
+    check = is_left
+
+    question_template = f"Does the camera face the left of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(60-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_right_object_center(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    
+    max_angle = 30
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[2], np.array([0,0,-1])), -1.0, 1.0))
+    is_right = angle_rad < max_angle / 180 * np.pi
+
+    check = is_right
+
+    question_template = f"Does the camera face the right of [A]?"
+    question = question_template.replace("[A]", A_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi-max_angle)/(60-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+
+
+def object_side_by_side_same_direction(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+
+    max_angle = 30
+
+
+    side_by_side_radius = np.abs(np.arctan(B_P_A[2]/ B_P_A[0]))
+    is_side_by_side = side_by_side_radius > (90 - max_angle) / 180 * np.pi
+
+    # 比较X轴的夹角
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_same_orientation = angle_rad < max_angle / 180 * np.pi
+    
+    check = is_same_orientation and is_side_by_side
+    
+    question_template = f"Is [A] and [B] side by side, facing the same direction?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 - np.abs(side_by_side_radius - (90 - max_angle) / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w2 = 1 - np.abs(angle_rad - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        score = 0 if w1 < 0 or w2 < 0 else w1 * w2
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score    
+
+def object_side_by_side_opposite_direction(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+
+    max_angle = 30
+    
+
+    side_by_side_radius = np.abs(np.arctan(B_P_A[2]/ B_P_A[0]))
+    is_side_by_side = side_by_side_radius > (90 - max_angle) / 180 * np.pi
+
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([-1,0,0])), -1.0, 1.0))
+    is_opposite_orientation = angle_rad < max_angle / 180 * np.pi
+
+    check = is_opposite_orientation and is_side_by_side
+
+    question_template = f"Is [A] and [B] side by side, facing the opposite direction?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 - np.abs(side_by_side_radius - (90 - max_angle) / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w2 = 1 - np.abs(angle_rad - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        score = 0 if w1 < 0 or w2 < 0 else w1 * w2
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_face_to_face(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+    
+    max_angle = 30
+    
+    
+    face_to_face_radius = np.abs(np.arctan(B_P_A[2]/ B_P_A[0]))
+    is_line =  B_P_A[0] > 0 and face_to_face_radius < max_angle * np.pi# 在一条线上，且A在B的前面
+    
+    
+    angle_rad = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], [-1,0,0]), -1.0, 1.0))
+    is_opposite_orientation = angle_rad < max_angle / 180 * np.pi
+    
+    check = is_opposite_orientation and is_line
+    
+    question_template = f"Is [A] and [B] face to face?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 if  B_P_A[0] > 0 else -1
+        w2 = 1 - np.abs(face_to_face_radius - max_angle * np.pi) / (np.pi / 12) # 15度的阈值
+        w3 = 1 - np.abs(angle_rad - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        score = 0 if w1 < 0 or w2 < 0 or w3 < 0 else w1 * w2 * w3
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+    
+def object_back_to_back(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"] 
+    B_P_A = B_rotation_matrix.T @ (A_pos - B_pos) # 在B物体参考系下，A物体的位置
+    A_rotation_matrix = B_rotation_matrix.T @ A["rotation_matrix"]
+
+    max_angle = 30
+
+    face_to_face_radius = np.abs(np.arctan(B_P_A[2]/ B_P_A[0]))
+    is_line =  B_P_A[0] < 0 and face_to_face_radius < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    angle_rad =  np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], [-1,0,0]), -1.0, 1.0))
+    is_opposite_orientation = angle_rad < max_angle / 180 * np.pi
+    
+    check = is_opposite_orientation and is_line
+    
+    question_template = f"Is [A] and [B] back to back?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 if  B_P_A[0] < 0 else -1
+        w2 = 1 - np.abs(face_to_face_radius - max_angle * np.pi) / (np.pi / 12) # 15度的阈值
+        w3 = 1 - np.abs(angle_rad - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        score = 0 if w1 < 0 or w2 < 0 or w3 < 0 else w1 * w2 * w3
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+
+
+def object_front(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    A_rotation_matrix = A["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，B物体的位置
+
+    max_angle = 15
+    A_P_B_direcetion = A_P_B / np.linalg.norm(A_P_B)
+    angle_rad = np.arccos(np.clip(np.dot(A_P_B_direcetion, np.array([1,0,0])), -1.0, 1.0))
+    
+    B_is_in_front_A = A_P_B[0] > 0 and angle_rad < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    check = B_is_in_front_A
+    
+    question_template = f"Is [B] in front of [A], from the view of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_back(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    A_rotation_matrix = A["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，B物体的位置
+
+    max_angle = 15
+    A_P_B_direcetion = A_P_B / np.linalg.norm(A_P_B)
+    angle_rad = np.arccos(np.clip(np.dot(A_P_B_direcetion, np.array([-1,0,0])), -1.0, 1.0))
+    B_is_in_back_A = A_P_B[0] < 0 and angle_rad < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    check = B_is_in_back_A
+    
+    question_template = f"Is [B] in back of [A], from the view of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (45-max_angle))
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def object_left(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    A_rotation_matrix = A["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，B物体的位置
+
+    max_angle = 30
+    A_P_B_direcetion = A_P_B / np.linalg.norm(A_P_B)
+    angle_rad = np.arccos(np.clip(np.dot(A_P_B_direcetion, np.array([0,0,-1])), -1.0, 1.0))
+    B_is_in_left_A = A_P_B[2] < 0 and angle_rad < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    check = B_is_in_left_A
+    
+    question_template = f"Is [B] on the left of [A], from the view of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (60-max_angle))
+        score = 0 if score < 0 or A_P_B[2] > 0 else score
+
+    return question, answer, check, score  
+
+def object_right(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # A_rotation_matrix = A["rotation_matrix"]
+    A_rotation_matrix = A["rotation_matrix"] 
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A物体参考系下，B物体的位置
+
+
+    max_angle = 30
+    A_P_B_direcetion = A_P_B / np.linalg.norm(A_P_B)
+    angle_rad = np.arccos(np.clip(np.dot(A_P_B_direcetion, np.array([0,0,1])), -1.0, 1.0))
+    B_is_in_right_A = A_P_B[2] > 0 and angle_rad < max_angle / 180 * np.pi# 在一条线上，且A在B的前面
+    
+    check = B_is_in_right_A
+    
+    question_template = f"Is [B] on the right of [A], from the view of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs((angle_rad*180/np.pi - max_angle) / (60-max_angle))
+        score = 0 if score < 0 or A_P_B[2] < 0 else score
+
+    return question, answer, check, score  
+
+
+
+def camera_two_objects_closer(A,B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # 计算距离
+    distance_a = np.linalg.norm(A_pos)
+    distance_b = np.linalg.norm(B_pos)
+    
+    
+    check = distance_a < distance_b
+
+    question_template = f"Is [A] closer to the camera than [B]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    score = 1 if check else 0
+    
+    return question, answer, check, score
+
+def camera_two_objects_farther(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # 计算距离
+    distance_a = np.linalg.norm(A_pos)
+    distance_b = np.linalg.norm(B_pos)
+    
+    check = distance_a > distance_b
+
+    question_template = f"Is [A] farther to the camera than [B]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    score = 1 if check else 0
+    
+    return question, answer, check, score
+
+def camera_two_objects_left(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    
+    check = B_pos[0] - A_pos[0] > 0 
+
+    question_template = f"Is [A] on the left of [B], from the view of the camera?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 1 if check else 0
+    
+    return question, answer, check, score
+
+def camera_two_objects_right(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    
+    check = B_pos[0] - A_pos[0] < 0 
+
+    question_template = f"Is [A] on the right of [B], from the view of the camera?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 1 if check else 0
+    
+    return question, answer, check, score
+
+
+
+def object_apart_0_5meter(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # 计算距离
+    distance = np.linalg.norm(A_pos - B_pos)
+    
+    delta = 1.0/3
+    gt_distance = 0.5
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] apart from [B] about 0.5 meter?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def object_apart_1meter(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # 计算距离
+    distance = np.linalg.norm(A_pos - B_pos)
+    
+    delta = 1.0/3
+    gt_distance = 1
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] apart from [B] about 1 meter?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_apart_1_5meter(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标��转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # 计算距离
+    distance = np.linalg.norm(A_pos - B_pos)
+    
+    delta = 1.0/3
+    gt_distance = 1.5
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] apart from [B] about 0.5 meter?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score
+
+def object_apart_2meter(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+
+    # 计算距离
+    distance = np.linalg.norm(A_pos - B_pos)
+    
+    delta = 1.0/3
+    gt_distance = 2
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] apart from [B] about 2 meters?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+
+def camera_1meter_away(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    delta = 1.0/3
+    gt_distance = 1
+
+    distance = np.linalg.norm(A_pos)
+
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is the camera about 1 meter away from [A]?"
+    question = question_template.replace("[A]", A_desc)
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_2meter_away(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    delta = 1.0/3
+    gt_distance = 2
+
+    distance = np.linalg.norm(A_pos)
+
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is the camera about 2 meter away from [A]?"
+    question = question_template.replace("[A]", A_desc)
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_3meter_away(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    delta = 0.3
+    gt_distance = 3
+
+    distance = np.linalg.norm(A_pos)
+
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is the camera about 3 meter away from [A]?"
+    question = question_template.replace("[A]", A_desc)
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def camera_4meter_away(A):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    
+    delta = 0.2
+    gt_distance = 4
+
+    distance = np.linalg.norm(A_pos)
+
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is the camera about 4 meter away from [A]?"
+    question = question_template.replace("[A]", A_desc)
+    answer = "Yes" if check else "No"
+
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+
+
+def object_bigger_than1_2(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 计算距离
+    A_rotation_matrix = A["rotation_matrix"]
+    theta_A = np.arctan2(A_rotation_matrix.T[0][2], A_rotation_matrix.T[0][0])
+    A_center = A["pcd"].get_center()
+    R = A["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_A)) 
+    A["pcd"] = A["pcd"].rotate(R)
+    A_length = A["pcd"].get_axis_aligned_bounding_box().get_extent()[0]
+    A_height = A["pcd"].get_axis_aligned_bounding_box().get_extent()[1]
+    A_width = A["pcd"].get_axis_aligned_bounding_box().get_extent()[2]
+    A_volume = A_length * A_height * A_width
+    
+    B_rotation_matrix = B["rotation_matrix"]
+    theta_B = np.arctan2(B_rotation_matrix.T[0][2], B_rotation_matrix.T[0][0])
+    B_center = B["pcd"].get_center()
+    R = B["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_B))
+    B["pcd"] = B["pcd"].rotate(R)
+    B_length = B["pcd"].get_axis_aligned_bounding_box().get_extent()[0]
+    B_height = B["pcd"].get_axis_aligned_bounding_box().get_extent()[1]
+    B_width = B["pcd"].get_axis_aligned_bounding_box().get_extent()[2]
+    B_volume = B_length * B_height * B_width
+    
+    if A_volume > B_volume:
+        distance = A_volume / B_volume
+    else:
+        distance = B_volume / A_volume
+    
+    delta = 1.0/3
+    gt_distance = 1.2
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] bigger than [B] about 0.2 times?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_higher_20cm(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 计算距离
+        
+    A_height = A["pcd"].get_axis_aligned_bounding_box().get_extent()[1]
+    B_height = B["pcd"].get_axis_aligned_bounding_box().get_extent()[1]
+    distance = np.abs(A_height-B_height)
+    
+    delta = 1.0/3
+    gt_distance = 0.2
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] higher 20cm than [B]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_longer_50cm(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 计算距离
+        
+    A_rotation_matrix = A["rotation_matrix"]
+    theta_A = np.arctan2(A_rotation_matrix.T[0][2], A_rotation_matrix.T[0][0])
+    A_center = A["pcd"].get_center()
+    R = A["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_A)) 
+    A["pcd"] = A["pcd"].rotate(R)
+    A_length = A["pcd"].get_axis_aligned_bounding_box().get_extent()[0]
+    
+    B_rotation_matrix = B["rotation_matrix"]
+    theta_B = np.arctan2(B_rotation_matrix.T[0][2], B_rotation_matrix.T[0][0])
+    B_center = B["pcd"].get_center()
+    R = B["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_B))
+    B["pcd"] = B["pcd"].rotate(R)
+    B_length = B["pcd"].get_axis_aligned_bounding_box().get_extent()[0]
+    
+    distance = np.abs(A_length-B_length)
+    
+    delta = 1.0/3
+    gt_distance = 0.5
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] longer 50cm than [B]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+def object_wider_30cm(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    A_desc, B_desc = A_desc.lower(), B_desc.lower()
+
+    # 计算距离
+        
+    A_rotation_matrix = A["rotation_matrix"]
+    theta_A = np.arctan2(A_rotation_matrix.T[0][2], A_rotation_matrix.T[0][0])
+    A_center = A["pcd"].get_center()
+    R = A["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_A)) 
+    A["pcd"] = A["pcd"].rotate(R)
+    A_width = A["pcd"].get_axis_aligned_bounding_box().get_extent()[2]
+    
+    B_rotation_matrix = B["rotation_matrix"]
+    theta_B = np.arctan2(B_rotation_matrix.T[0][2], B_rotation_matrix.T[0][0])
+    B_center = B["pcd"].get_center()
+    R = B["pcd"].get_rotation_matrix_from_xyz((0, 0, theta_B))
+    B["pcd"] = B["pcd"].rotate(R)
+    B_width = B["pcd"].get_axis_aligned_bounding_box().get_extent()[2]
+    
+    distance = np.abs(A_width-B_width)
+    
+    delta = 1.0/3
+    gt_distance = 0.3
+    
+    check = (1-delta)*gt_distance < distance and distance < (1+delta)*gt_distance
+
+    question_template = f"Is [A] wider 30cm than [B]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        score = 1 - 1*np.abs(((distance - gt_distance) / gt_distance)- delta)/delta
+        score = 0 if score < 0 else score
+
+    return question, answer, check, score  
+
+
+def side_by_side_front(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    max_angle = 30
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    side_by_side_radius = np.abs(np.arctan(A_P_B[2]/ A_P_B[0]))
+    is_side_by_side = side_by_side_radius > (90 - max_angle) * np.pi / 180
+    
+    same_direction_radius = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_same_direction = same_direction_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    front_radius = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,-1])), -1.0, 1.0))
+    is_front = front_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    check = is_side_by_side and is_same_direction and is_front
+    
+    question_template = f"Is [A] and [B] side-by-side and same-orientation with viewed from the front of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 - np.abs(side_by_side_radius - (90 - max_angle) / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w2 = 1 - np.abs(same_direction_radius - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w3 = 1 - np.abs(front_radius - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+
+    return question, answer, check, score
+
+def side_by_side_left(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    max_angle = 30
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    side_by_side_radius = np.abs(np.arctan(A_P_B[2]/ A_P_B[0]))
+    is_side_by_side = side_by_side_radius > (90 - max_angle) * np.pi / 180
+    
+    same_direction_radius = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_same_direction = same_direction_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    left_radius = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([-1,0,0])), -1.0, 1.0))
+    is_left = left_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    check = is_side_by_side and is_same_direction and is_left
+    
+    question_template = f"Is [A] and [B] side-by-side and same-orientation with viewed from the left of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 - np.abs(side_by_side_radius - (90 - max_angle) / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w2 = 1 - np.abs(same_direction_radius - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w3 = 1 - np.abs(left_radius - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+
+    return question, answer, check, score
+
+def side_by_side_right(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    max_angle = 30
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    side_by_side_radius = np.abs(np.arctan(A_P_B[2]/ A_P_B[0]))
+    is_side_by_side = side_by_side_radius > (90 - max_angle) * np.pi / 180
+    
+    same_direction_radius = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_same_direction = same_direction_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    right_radius = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_right = right_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    check = is_side_by_side and is_same_direction and is_right
+    
+    question_template = f"Is [A] and [B] side-by-side and same-orientation with viewed from the right of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 - np.abs(side_by_side_radius - (90 - max_angle) / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w2 = 1 - np.abs(same_direction_radius - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w3 = 1 - np.abs(right_radius - max_angle / 180 * np.pi) / (np.pi / 6) # 30度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+
+    return question, answer, check, score
+
+
+def side_by_side_back(A, B):
+    A_desc, A_cloud = A["caption"], A["pcd"]
+    A_desc = A_desc.lower()
+    B_desc, B_cloud = B["caption"], B["pcd"]
+    B_desc = B_desc.lower()
+    
+    # 从PyTorch3D的坐标系转换到OpenCV的坐标系
+    A_pos = A_cloud.get_center()
+    A_pos[0] = -A_pos[0]; A_pos[1] = -A_pos[1]
+    B_pos = B_cloud.get_center()
+    B_pos[0] = -B_pos[0]; B_pos[1] = -B_pos[1]
+    
+    
+    A_rotation_matrix = A["rotation_matrix"]
+    B_rotation_matrix = B["rotation_matrix"]
+    
+    B_rotation_matrix = A_rotation_matrix.T @ B_rotation_matrix # 在A的坐标系下，B的旋转矩阵
+    
+    max_angle = 30
+    
+    A_P_B = A_rotation_matrix.T @ (B_pos - A_pos) # 在A的坐标系下，B相对于A的位置
+    side_by_side_radius = np.abs(np.arctan(A_P_B[2]/ A_P_B[0]))
+    is_side_by_side = side_by_side_radius > (90 - max_angle) * np.pi / 180
+    
+    same_direction_radius = np.arccos(np.clip(np.dot(B_rotation_matrix.T[0], np.array([1,0,0])), -1.0, 1.0))
+    is_same_direction = same_direction_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    back_radius = np.arccos(np.clip(np.dot(A_rotation_matrix.T[0], np.array([0,0,1])), -1.0, 1.0))
+    is_back = back_radius < max_angle * np.pi / 180 # 30度的阈值
+    
+    check = is_side_by_side and is_same_direction and is_back
+    
+    question_template = f"Is [A] and [B] side-by-side and same-orientation with viewed from the back of [A]?"
+    question = question_template.replace("[A]", A_desc).replace("[B]", B_desc)
+    
+    answer = "Yes" if check else "No"
+    
+    score = 0
+    if check:
+        score = 1
+    else:
+        w1 = 1 - np.abs(side_by_side_radius - (90 - max_angle) / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w2 = 1 - np.abs(same_direction_radius - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        w3 = 1 - np.abs(back_radius - max_angle / 180 * np.pi) / (np.pi / 12) # 15度的阈值
+        score = 0 if w1<0 or w2<0 or w3<0 else w1 * w2 * w3
+
+    return question, answer, check, score

processor/prompt_utils.py

@@ -0,0 +1,131 @@
+import math
+import random
+import string
+
+import numpy as np
+
+
+def generate_random_string(length):
+    letters = string.ascii_letters + string.digits
+    return "".join(random.choice(letters) for _ in range(length))
+
+
+def calculate_angle_clockwise(A_pos, B_pos, x_right=False):
+    # Vector from A to B
+    if x_right:
+        vector_A_to_B = (A_pos[0] - B_pos[0], B_pos[1] - A_pos[1])
+    else:
+        vector_A_to_B = (B_pos[0] - A_pos[0], B_pos[1] - A_pos[1])
+
+    # Angle of this vector w.r.t. positive z-axis
+    angle_rad = math.atan2(vector_A_to_B[0], vector_A_to_B[1])  # atan2 handles all quadrants
+    angle_deg = math.degrees(angle_rad)
+
+    # Convert angle to clock position, 360 degrees => 12 hours, so 1 hour = 30 degrees
+    # We adjust the angle to be positive and then calculate the clock position
+    angle_deg = (angle_deg + 360) % 360
+    clock_position = 12 - angle_deg // 30
+    clock_position = clock_position if clock_position > 0 else 12 + clock_position
+
+    return clock_position
+
+
+def is_aligned_vertically(A, B):
+    # Convert Open3D point cloud to NumPy array
+    A_points = np.asarray(A["pcd"].points)
+    B_points = np.asarray(B["pcd"].points)
+
+    # Calculate vertical (y) extents for A
+    A_min, A_max = np.min(A_points[:, 1]), np.max(A_points[:, 1])
+    # Calculate vertical (y) extents for B
+    B_min, B_max = np.min(B_points[:, 1]), np.max(B_points[:, 1])
+
+    # Determine vertical overlap
+    overlap = max(0, min(A_max, B_max) - max(A_min, B_min))
+    A_overlap_percentage = overlap / (A_max - A_min) if A_max != A_min else 0
+    B_overlap_percentage = overlap / (B_max - B_min) if B_max != B_min else 0
+
+    # Return True if both overlaps are greater than 50%
+    return A_overlap_percentage > 0.5 and B_overlap_percentage > 0.5
+
+
+def is_aligned_horizontally(A, B):
+    # The high level logic is to check if the x-axis of one object is fully contained by the x-axis of the other object
+
+    # Extract the bounding boxes for both A and B
+    A_box = A["pcd"].get_axis_aligned_bounding_box()
+    B_box = B["pcd"].get_axis_aligned_bounding_box()
+
+    # Get the min and max x-axis values for both A and B
+    A_min_x, A_max_x = A_box.get_min_bound()[0], A_box.get_max_bound()[0]
+    B_min_x, B_max_x = B_box.get_min_bound()[0], B_box.get_max_bound()[0]
+
+    # Check if A and B are almost the same size on the x-axis
+    A_width, B_width = A_max_x - A_min_x, B_max_x - B_min_x
+    is_almost_same_size = max(A_width, B_width) / min(A_width, B_width) <= 1.5
+    if not is_almost_same_size:
+        return False
+
+    overlap_min, overlap_max = max(A_min_x, B_min_x), min(A_max_x, B_max_x)
+    overlap_width = max(0, overlap_max - overlap_min)
+    overlap_percent = max(overlap_width / A_width, overlap_width / B_width)
+
+    return overlap_percent > 0.95
+
+
+def is_y_axis_overlapped(A, B):
+    # Extract the y-axis values (height) of the bounding boxes
+    A_box = A["pcd"].get_axis_aligned_bounding_box()
+    B_box = B["pcd"].get_axis_aligned_bounding_box()
+
+    # Get the min and max y-axis values for both A and B
+    A_min_y, A_max_y = A_box.get_min_bound()[1], A_box.get_max_bound()[1]
+    B_min_y, B_max_y = B_box.get_min_bound()[1], B_box.get_max_bound()[1]
+
+    # Check if there's any overlap in the y-axis values
+    # There are four possible scenarios for overlap, but we can check them with a simpler logic:
+    # If one box's minimum is between the other's min and max, or one box's max is.
+    overlap = (A_min_y <= B_max_y and A_max_y >= B_min_y) or (B_min_y <= A_max_y and B_max_y >= A_min_y)
+
+    return overlap
+
+
+def is_supporting(A, B):
+    # Extract bounding boxes
+    A_box = A["pcd"].get_axis_aligned_bounding_box()
+    B_box = B["pcd"].get_axis_aligned_bounding_box()
+
+    # Get the corners of the bounding boxes
+    A_min, A_max = A_box.get_min_bound(), A_box.get_max_bound()
+    B_min, B_max = B_box.get_min_bound(), B_box.get_max_bound()
+
+    # Check vertical contact:
+    # The bottom of the upper object is at or above the top of the lower object
+    vertical_contact = (A_min[2] <= B_max[2] and A_min[2] >= B_min[2]) or (
+        B_min[2] <= A_max[2] and B_min[2] >= A_min[2]
+    )
+
+    if not vertical_contact:
+        # If there's no vertical contact, they are not supporting each other
+        return False
+
+    # Determine which object is on top and which is on bottom
+    if A_min[2] < B_min[2]:
+        top, bottom = B, A
+        top_min, top_max = B_min, B_max
+        bottom_min, bottom_max = A_min, A_max
+    else:
+        top, bottom = A, B
+        top_min, top_max = A_min, A_max
+        bottom_min, bottom_max = B_min, B_max
+
+    # Check horizontal coverage:
+    # The larger (top) object's bounding box completely covers the smaller (bottom) object's bounding box
+    horizontal_coverage = (
+        top_min[0] <= bottom_min[0]
+        and top_max[0] >= bottom_max[0]
+        and top_min[1] <= bottom_min[1]
+        and top_max[1] >= bottom_max[1]
+    )
+
+    return horizontal_coverage

processor/segment.py

@@ -0,0 +1,238 @@
+import cv2
+import torch
+import torchvision
+from osdsynth.processor.wrappers.grounding_dino import get_grounding_dino_model
+from osdsynth.processor.wrappers.ram import get_tagging_model, run_tagging_model
+from osdsynth.processor.wrappers.sam import (
+    convert_detections_to_dict,
+    convert_detections_to_list,
+    crop_detections_with_xyxy,
+    filter_detections,
+    get_sam_predictor,
+    get_sam_segmentation_from_xyxy,
+    mask_subtract_contained,
+    post_process_mask,
+    sort_detections_by_area,
+)
+from osdsynth.utils.logger import SkipImageException
+from osdsynth.visualizer.som import draw_som_on_image
+from PIL import Image
+import numpy as np
+
+class SegmentImage:
+    """Class to segment the image."""
+
+    def __init__(self, cfg, logger, device, init_gdino=True, init_tagging=True, init_sam=True):
+        self.cfg = cfg
+        self.logger = logger
+        self.device = device
+
+        if init_gdino:
+            # Initialize the Grounding Dino Model
+            self.grounding_dino_model = get_grounding_dino_model(cfg, device)
+        else:
+            self.grounding_dino_model = None
+
+        if init_tagging:
+            # Initialize the tagging Model
+            self.tagging_transform, self.tagging_model = get_tagging_model(cfg, device)
+        else:
+            self.tagging_transform = self.tagging_model = None
+
+        if init_sam:
+            # Initialize the SAM Model
+            self.sam_predictor = get_sam_predictor(cfg.sam_variant, device)
+        else:
+            self.sam_predictor = None
+
+        pass
+
+    def process(self, image_bgr, two_class ,plot_som=True):
+        """Segment the image."""
+
+        image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
+        image_rgb_pil = Image.fromarray(image_rgb)
+        
+        # image_rgb_pil.save('tmp.png')
+
+        img_tagging = image_rgb_pil.resize((384, 384))
+        img_tagging = self.tagging_transform(img_tagging).unsqueeze(0).to(self.device)
+
+        # Tag2Text
+        if two_class is None:    
+            classes = run_tagging_model(self.cfg, img_tagging, self.tagging_model)
+        else:
+            classes = two_class
+
+        if len(classes) == 0:
+            raise SkipImageException("No foreground objects detected by tagging model.")
+
+        # Using GroundingDINO to detect and SAM to segment
+        detections = self.grounding_dino_model.predict_with_classes(
+            image=image_bgr,  # This function expects a BGR image...
+            classes=classes,
+            box_threshold=self.cfg.box_threshold,
+            text_threshold=self.cfg.text_threshold,
+        )
+
+
+        if len(detections.class_id) < 1:
+            raise SkipImageException("No object detected.")
+        
+
+
+        # Non-maximum suppression
+        nms_idx = (
+            torchvision.ops.nms(
+                torch.from_numpy(detections.xyxy),
+                torch.from_numpy(detections.confidence),
+                self.cfg.nms_threshold,
+            )
+            .numpy()
+            .tolist()
+        )
+
+        print(f"Before NMS: {len(detections.xyxy)} detections")
+        detections.xyxy = detections.xyxy[nms_idx]
+        detections.confidence = detections.confidence[nms_idx]
+        detections.class_id = detections.class_id[nms_idx]
+        print(f"After NMS: {len(detections.xyxy)} detections")
+
+        # Somehow some detections will have class_id=-1, remove them
+        valid_idx = detections.class_id != -1
+        detections.xyxy = detections.xyxy[valid_idx]
+        detections.confidence = detections.confidence[valid_idx]
+        detections.class_id = detections.class_id[valid_idx]
+
+        # Segment Anything
+        detections.mask = get_sam_segmentation_from_xyxy(
+            sam_predictor=self.sam_predictor, image=image_rgb, xyxy=detections.xyxy
+        )
+
+        # Convert the detection to a dict. Elements are np.ndarray
+        detections_dict = convert_detections_to_dict(detections, classes)
+        
+        # Filter out the objects based on various criteria
+        detections_dict = filter_detections(self.cfg, detections_dict, image_rgb)
+
+        if len(detections_dict["xyxy"]) < 1:
+            raise SkipImageException("No object detected after filtering.")
+
+        # Subtract the mask of bounding boxes that are contained by it
+        detections_dict["subtracted_mask"], mask_contained = mask_subtract_contained(
+            detections_dict["xyxy"], detections_dict["mask"], th1=0.05, th2=0.05
+        )
+
+        # Sort the dets by area
+        detections_dict = sort_detections_by_area(detections_dict)
+
+        # Add RLE to dict
+        detections_dict = post_process_mask(detections_dict)
+
+        # Convert the detection to a list. Each element is a dict
+        detections_list = convert_detections_to_list(detections_dict, classes)
+        
+        # Skip objects with confidence lower than 0.4
+        # detections_list = skipbyconfidence(detections_list)
+
+        detections_list = crop_detections_with_xyxy(self.cfg, image_rgb_pil, detections_list)
+
+        detections_list = segmentImage(detections_list, image_rgb_pil)
+        
+        detections_list = add_index_to_class(detections_list)
+        
+        if two_class is not None:
+            if len(two_class)==2 and len(detections_list) != 2:
+                raise SkipImageException("Not all objects detected.")
+            
+            if len(two_class)==1 and len(detections_list) != 1:
+                raise SkipImageException("Not all objects detected.")
+            
+            if len(two_class)==3 and len(detections_list) != 3:
+                raise SkipImageException("Not all objects detected.")
+            
+            if len(two_class)==2:
+                detections_two_class = [detections_list[0]['class_name'][:-1], detections_list[1]['class_name'][:-1]]
+                if two_class[0] not in detections_two_class or two_class[1] not in detections_two_class:
+                    raise SkipImageException("Not all objects detected.")
+                
+            if len(two_class)==3:
+                detections_two_class = [detections_list[0]['class_name'][:-1], detections_list[1]['class_name'][:-1], detections_list[2]['class_name'][:-1]]
+                if two_class[0] not in detections_two_class or two_class[1] not in detections_two_class or two_class[2] not in detections_two_class:
+                    raise SkipImageException("Not all objects detected.")
+            
+        
+
+        if plot_som:
+            # Visualize with SoM
+            vis_som = draw_som_on_image(
+                detections_dict,
+                image_rgb,
+                label_mode="1",
+                alpha=0.4,
+                anno_mode=["Mask", "Mark", "Box"],
+            )
+        else:
+            vis_som = None
+        
+        
+            
+        return vis_som, detections_list
+
+#  Copy the object area from the original image to a transparent background
+def segmentImage(detections_list, image_rgb_pil):
+    
+    for i in range(len(detections_list)):
+        image_pil = detections_list[i]['image_crop']
+        mask_pil = Image.fromarray(detections_list[i]['mask_crop'])
+        
+        image_rgba = image_pil.convert("RGBA")
+        
+        transparent_bg = Image.new("RGBA", image_rgba.size, (0, 0, 0, 0))
+
+        # Copy the object area from the original image to a transparent background using a mask
+        segmented_image = Image.composite(
+            image_rgba,       
+            transparent_bg,
+            mask_pil          
+        )
+        
+        detections_list[i]['image_segment'] = segmented_image
+    
+    return detections_list
+    
+def skipbyconfidence(detections_list):
+    skip_index = []
+    for i in range(len(detections_list)):
+        if detections_list[i]['confidence'] < 0.3:
+            skip_index.append(i)
+    
+    for i in skip_index[::-1]:
+        del detections_list[i]
+    
+    return detections_list
+    
+def add_bbox_and_taggingtext_to_image(image, detections_list):
+    for i in range(len(detections_list)):
+        bbox = detections_list[i]['xyxy']
+        label = detections_list[i]['class_name']
+        confidence = detections_list[i]['confidence']
+        
+        cv2.rectangle(image, (int(bbox[0]), int(bbox[1])), (int(bbox[2]), int(bbox[3])), (0, 255, 0), 2)
+        cv2.putText(image, f"{label} {confidence:.2f}", (int(bbox[0]), int((bbox[1]+bbox[3])/2)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)
+    
+    return image
+
+def add_index_to_class(detections_list):
+    # If a class appears for the first time, add 0 to the object's class_name, add 1 to the second appearance, and so on
+    class_index = {}
+    for detection in detections_list:
+        class_name = detection['class_name']
+        if class_name not in class_index:
+            class_index[class_name] = 0
+        else:
+            class_index[class_name] += 1
+
+        detection['class_name'] = f"{class_name}{class_index[class_name]}"
+    return detections_list
+        

utils/logger.py

@@ -0,0 +1,159 @@
+import atexit
+import json
+import logging
+import os
+import sys
+
+import numpy as np
+import open3d as o3d
+from termcolor import colored
+
+__all__ = [
+    "setup_logger",
+]
+
+
+def save_detection_list_to_json(detection_list, output_file):
+    def serialize_numpy(obj):
+        if isinstance(obj, np.ndarray):
+            return obj.tolist()
+        return obj
+
+    def serialize_open3d(obj):
+        if isinstance(obj, o3d.geometry.AxisAlignedBoundingBox):
+            return {
+                "type": "AxisAlignedBoundingBox",
+                "min_bound": obj.min_bound.tolist(),
+                "max_bound": obj.max_bound.tolist(),
+            }
+        elif isinstance(obj, o3d.geometry.OrientedBoundingBox):
+            return {
+                "type": "OrientedBoundingBox",
+                "center": obj.center.tolist(),
+                "extent": obj.extent.tolist(),
+                "R": obj.R.tolist(),
+            }
+        elif isinstance(obj, o3d.geometry.PointCloud):
+            return {
+                "type": "PointCloud",
+                "points": np.asarray(obj.points).tolist(),
+                "colors": np.asarray(obj.colors).tolist() if obj.has_colors() else None,
+                "normals": np.asarray(obj.normals).tolist() if obj.has_normals() else None,
+            }
+        return obj
+
+    def serialize_detection(detection):
+        serialized = {}
+        for key, value in detection.items():
+            if key in ["axis_aligned_bbox", "oriented_bbox", "pcd"]:
+                serialized[key] = serialize_open3d(value)
+            elif isinstance(value, np.ndarray):
+                serialized[key] = serialize_numpy(value)
+            elif isinstance(value, (list, dict, str, int, float, bool, type(None))):
+                serialized[key] = value
+            else:
+                serialized[key] = str(value)  # Convert other types to string
+        return serialized
+
+    serialized_list = [serialize_detection(detection) for detection in detection_list]
+
+    with open(output_file, "w") as f:
+        json.dump(serialized_list, f, indent=2)
+
+    print(f"Detection list saved to {output_file}")
+
+
+class SkipImageException(Exception):
+    def __init__(self, message="Known exception, skip the image."):
+        # Call the base class constructor with the parameters it needs
+        super().__init__(message)
+
+
+class _ColorfulFormatter(logging.Formatter):
+    def __init__(self, *args, **kwargs):
+        self._root_name = kwargs.pop("root_name") + "."
+        self._abbrev_name = kwargs.pop("abbrev_name", "")
+        if len(self._abbrev_name):
+            self._abbrev_name = self._abbrev_name + "."
+        super().__init__(*args, **kwargs)
+
+    def formatMessage(self, record):
+        record.name = record.name.replace(self._root_name, self._abbrev_name)
+        log = super().formatMessage(record)
+        if record.levelno == logging.WARNING:
+            prefix = colored("WARNING", "red", attrs=["blink"])
+        elif record.levelno == logging.ERROR or record.levelno == logging.CRITICAL:
+            prefix = colored("ERROR", "red", attrs=["blink", "underline"])
+        else:
+            return log
+        return prefix + " " + log
+
+
+def setup_logger(output=None, distributed_rank=0, *, name="metricdepth", color=True, abbrev_name=None):
+    """Initialize the detectron2 logger and set its verbosity level to "DEBUG".
+
+    Args:
+        output (str): a file name or a directory to save log. If None, will not save log file.
+            If ends with ".txt" or ".log", assumed to be a file name.
+            Otherwise, logs will be saved to `output/log.txt`.
+        abbrev_name (str): an abbreviation of the module, to avoid log names in logs.
+            Set to "" not log the root module in logs.
+            By default, will abbreviate "detectron2" to "d2" and leave other
+            modules unchanged.
+    Returns:
+        logging.Logger: a logger
+    """
+    logger = logging.getLogger()
+    logger.setLevel(logging.INFO)  # NOTE: if more detailed, change it to logging.DEBUG
+    logger.propagate = False
+
+    if abbrev_name is None:
+        abbrev_name = "d2"
+
+    plain_formatter = logging.Formatter("[%(asctime)s] %(name)s %(levelname)s %(message)s ", datefmt="%m/%d %H:%M:%S")
+    # stdout logging: master  only
+    if distributed_rank == 0:
+        ch = logging.StreamHandler(stream=sys.stdout)
+        ch.setLevel(logging.INFO)  # NOTE: if more detailed, change it to logging.DEBUG
+        if color:
+            formatter = _ColorfulFormatter(
+                colored("[%(asctime)s %(name)s]: ", "green") + "%(message)s",
+                datefmt="%m/%d %H:%M:%S",
+                root_name=name,
+                abbrev_name=str(abbrev_name),
+            )
+        else:
+            formatter = plain_formatter
+        ch.setFormatter(formatter)
+        logger.addHandler(ch)
+
+    # file logging: all workers
+    if output is not None:
+        if output.endswith(".txt") or output.endswith(".log"):
+            filename = output
+        else:
+            filename = os.path.join(output, "log.txt")
+        if distributed_rank > 0:
+            filename = filename + f".rank{distributed_rank}"
+        os.makedirs(os.path.dirname(filename), exist_ok=True)
+
+        fh = logging.StreamHandler(_cached_log_stream(filename))
+        fh.setLevel(logging.INFO)  # NOTE: if more detailed, change it to logging.DEBUG
+        fh.setFormatter(plain_formatter)
+        logger.addHandler(fh)
+
+    return logger
+
+
+from iopath.common.file_io import PathManager as PathManagerBase
+
+PathManager = PathManagerBase()
+
+
+# cache the opened file object, so that different calls to 'setup_logger
+# with the same file name can safely write to the same file.
+def _cached_log_stream(filename):
+    # use 1K buffer if writting to cloud storage
+    io = PathManager.open(filename, "a", buffering=1024 if "://" in filename else -1)
+    atexit.register(io.close)
+    return io

visualizer/som.py

@@ -0,0 +1,1429 @@
+# Copyright (c) Facebook, Inc. and its affiliates.
+import colorsys
+import logging
+import math
+import random
+from enum import Enum, unique
+
+import cv2
+import matplotlib as mpl
+import matplotlib.colors as mplc
+import matplotlib.figure as mplfigure
+import numpy as np
+import pycocotools.mask as mask_util
+import torch
+from detectron2.data import MetadataCatalog
+from detectron2.structures import BitMasks, Boxes, BoxMode, Keypoints, PolygonMasks, RotatedBoxes
+from detectron2.utils.colormap import random_color
+from detectron2.utils.file_io import PathManager
+from matplotlib.backends.backend_agg import FigureCanvasAgg
+from PIL import Image
+
+logger = logging.getLogger(__name__)
+
+__all__ = ["ColorMode", "VisImage", "Visualizer"]
+
+_SMALL_OBJECT_AREA_THRESH = 1000
+_LARGE_MASK_AREA_THRESH = 120000
+_OFF_WHITE = (1.0, 1.0, 240.0 / 255)
+_BLACK = (0, 0, 0)
+_RED = (1.0, 0, 0)
+
+_KEYPOINT_THRESHOLD = 0.05
+
+import numpy as np
+from detectron2.data import MetadataCatalog
+from PIL import Image
+
+metadata = MetadataCatalog.get("coco_2017_train_panoptic")
+
+
+def draw_mask_and_number_on_image(image_rgb, masks, labels, label_mode="1", alpha=0.1, anno_mode=["Mask"]):
+    img_width, img_height = image_rgb.size
+    visual = Visualizer(image_rgb, metadata=metadata)
+    for i in range(len(masks)):
+        mask = masks[i]
+        if mask.shape[0] != img_height or mask.shape[1] != img_width:
+            # Resize mask using PIL
+            mask_pil = Image.fromarray(mask)
+            mask_resized_pil = mask_pil.resize((img_width, img_height), Image.NEAREST)
+            mask = np.array(mask_resized_pil)
+
+        demo = visual.draw_binary_mask_with_number(
+            mask,
+            text=str(labels[i]),
+            label_mode=label_mode,
+            alpha=alpha,
+            anno_mode=anno_mode,
+        )
+    im = demo.get_image()
+    return im
+
+
+def draw_som_on_image(detections_dict, image_rgb, label_mode="1", alpha=0.1, anno_mode=["Mask"]):
+    # img_width, img_height = image_rgb.size
+    visual = Visualizer(image_rgb, metadata=metadata)
+    label = 0
+    mask_map = np.zeros(image_rgb.shape, dtype=np.uint8)
+    for i in range(len(detections_dict["xyxy"])):
+        mask = detections_dict["mask"][i]
+        demo = visual.draw_binary_mask_with_number(
+            mask,
+            text=str(label),
+            label_mode=label_mode,
+            alpha=alpha,
+            anno_mode=anno_mode,
+        )
+        mask_map[mask == 1] = label
+        label += 1
+    im = demo.get_image()
+    return im
+
+
+@unique
+class ColorMode(Enum):
+    """Enum of different color modes to use for instance visualizations."""
+
+    IMAGE = 0
+    """Picks a random color for every instance and overlay segmentations with low opacity."""
+    SEGMENTATION = 1
+    """Let instances of the same category have similar colors (from metadata.thing_colors), and overlay them with high
+    opacity.
+
+    This provides more attention on the quality of segmentation.
+    """
+    IMAGE_BW = 2
+    """Same as IMAGE, but convert all areas without masks to gray-scale.
+
+    Only available for drawing per-instance mask predictions.
+    """
+
+
+class GenericMask:
+    """
+    Attribute:
+        polygons (list[ndarray]): list[ndarray]: polygons for this mask.
+            Each ndarray has format [x, y, x, y, ...]
+        mask (ndarray): a binary mask
+    """
+
+    def __init__(self, mask_or_polygons, height, width):
+        self._mask = self._polygons = self._has_holes = None
+        self.height = height
+        self.width = width
+
+        m = mask_or_polygons
+        if isinstance(m, dict):
+            # RLEs
+            assert "counts" in m and "size" in m
+            if isinstance(m["counts"], list):  # uncompressed RLEs
+                h, w = m["size"]
+                assert h == height and w == width
+                m = mask_util.frPyObjects(m, h, w)
+            self._mask = mask_util.decode(m)[:, :]
+            return
+
+        if isinstance(m, list):  # list[ndarray]
+            self._polygons = [np.asarray(x).reshape(-1) for x in m]
+            return
+
+        if isinstance(m, np.ndarray):  # assumed to be a binary mask
+            assert m.shape[1] != 2, m.shape
+            assert m.shape == (
+                height,
+                width,
+            ), f"mask shape: {m.shape}, target dims: {height}, {width}"
+            self._mask = m.astype("uint8")
+            return
+
+        raise ValueError(f"GenericMask cannot handle object {m} of type '{type(m)}'")
+
+    @property
+    def mask(self):
+        if self._mask is None:
+            self._mask = self.polygons_to_mask(self._polygons)
+        return self._mask
+
+    @property
+    def polygons(self):
+        if self._polygons is None:
+            self._polygons, self._has_holes = self.mask_to_polygons(self._mask)
+        return self._polygons
+
+    @property
+    def has_holes(self):
+        if self._has_holes is None:
+            if self._mask is not None:
+                self._polygons, self._has_holes = self.mask_to_polygons(self._mask)
+            else:
+                self._has_holes = False  # if original format is polygon, does not have holes
+        return self._has_holes
+
+    def mask_to_polygons(self, mask):
+        # cv2.RETR_CCOMP flag retrieves all the contours and arranges them to a 2-level
+        # hierarchy. External contours (boundary) of the object are placed in hierarchy-1.
+        # Internal contours (holes) are placed in hierarchy-2.
+        # cv2.CHAIN_APPROX_NONE flag gets vertices of polygons from contours.
+        mask = np.ascontiguousarray(mask)  # some versions of cv2 does not support incontiguous arr
+        res = cv2.findContours(mask.astype("uint8"), cv2.RETR_CCOMP, cv2.CHAIN_APPROX_NONE)
+        hierarchy = res[-1]
+        if hierarchy is None:  # empty mask
+            return [], False
+        has_holes = (hierarchy.reshape(-1, 4)[:, 3] >= 0).sum() > 0
+        res = res[-2]
+        res = [x.flatten() for x in res]
+        # These coordinates from OpenCV are integers in range [0, W-1 or H-1].
+        # We add 0.5 to turn them into real-value coordinate space. A better solution
+        # would be to first +0.5 and then dilate the returned polygon by 0.5.
+        res = [x + 0.5 for x in res if len(x) >= 6]
+        return res, has_holes
+
+    def polygons_to_mask(self, polygons):
+        rle = mask_util.frPyObjects(polygons, self.height, self.width)
+        rle = mask_util.merge(rle)
+        return mask_util.decode(rle)[:, :]
+
+    def area(self):
+        return self.mask.sum()
+
+    def bbox(self):
+        p = mask_util.frPyObjects(self.polygons, self.height, self.width)
+        p = mask_util.merge(p)
+        bbox = mask_util.toBbox(p)
+        bbox[2] += bbox[0]
+        bbox[3] += bbox[1]
+        return bbox
+
+
+class _PanopticPrediction:
+    """Unify different panoptic annotation/prediction formats."""
+
+    def __init__(self, panoptic_seg, segments_info, metadata=None):
+        if segments_info is None:
+            assert metadata is not None
+            # If "segments_info" is None, we assume "panoptic_img" is a
+            # H*W int32 image storing the panoptic_id in the format of
+            # category_id * label_divisor + instance_id. We reserve -1 for
+            # VOID label.
+            label_divisor = metadata.label_divisor
+            segments_info = []
+            for panoptic_label in np.unique(panoptic_seg.numpy()):
+                if panoptic_label == -1:
+                    # VOID region.
+                    continue
+                pred_class = panoptic_label // label_divisor
+                isthing = pred_class in metadata.thing_dataset_id_to_contiguous_id.values()
+                segments_info.append(
+                    {
+                        "id": int(panoptic_label),
+                        "category_id": int(pred_class),
+                        "isthing": bool(isthing),
+                    }
+                )
+        del metadata
+
+        self._seg = panoptic_seg
+
+        self._sinfo = {s["id"]: s for s in segments_info}  # seg id -> seg info
+        segment_ids, areas = torch.unique(panoptic_seg, sorted=True, return_counts=True)
+        areas = areas.numpy()
+        sorted_idxs = np.argsort(-areas)
+        self._seg_ids, self._seg_areas = segment_ids[sorted_idxs], areas[sorted_idxs]
+        self._seg_ids = self._seg_ids.tolist()
+        for sid, area in zip(self._seg_ids, self._seg_areas):
+            if sid in self._sinfo:
+                self._sinfo[sid]["area"] = float(area)
+
+    def non_empty_mask(self):
+        """
+        Returns:
+            (H, W) array, a mask for all pixels that have a prediction
+        """
+        empty_ids = []
+        for id in self._seg_ids:
+            if id not in self._sinfo:
+                empty_ids.append(id)
+        if len(empty_ids) == 0:
+            return np.zeros(self._seg.shape, dtype=np.uint8)
+        assert len(empty_ids) == 1, ">1 ids corresponds to no labels. This is currently not supported"
+        return (self._seg != empty_ids[0]).numpy().astype(np.bool)
+
+    def semantic_masks(self):
+        for sid in self._seg_ids:
+            sinfo = self._sinfo.get(sid)
+            if sinfo is None or sinfo["isthing"]:
+                # Some pixels (e.g. id 0 in PanopticFPN) have no instance or semantic predictions.
+                continue
+            yield (self._seg == sid).numpy().astype(np.bool), sinfo
+
+    def instance_masks(self):
+        for sid in self._seg_ids:
+            sinfo = self._sinfo.get(sid)
+            if sinfo is None or not sinfo["isthing"]:
+                continue
+            mask = (self._seg == sid).numpy().astype(np.bool)
+            if mask.sum() > 0:
+                yield mask, sinfo
+
+
+def _create_text_labels(classes, scores, class_names, is_crowd=None):
+    """
+    Args:
+        classes (list[int] or None):
+        scores (list[float] or None):
+        class_names (list[str] or None):
+        is_crowd (list[bool] or None):
+
+    Returns:
+        list[str] or None
+    """
+    labels = None
+    if classes is not None:
+        if class_names is not None and len(class_names) > 0:
+            labels = [class_names[i] for i in classes]
+        else:
+            labels = [str(i) for i in classes]
+    if scores is not None:
+        if labels is None:
+            labels = [f"{s * 100:.0f}%" for s in scores]
+        else:
+            labels = [f"{l} {s * 100:.0f}%" for l, s in zip(labels, scores)]
+    if labels is not None and is_crowd is not None:
+        labels = [l + ("|crowd" if crowd else "") for l, crowd in zip(labels, is_crowd)]
+    return labels
+
+
+class VisImage:
+    def __init__(self, img, scale=1.0):
+        """
+        Args:
+            img (ndarray): an RGB image of shape (H, W, 3) in range [0, 255].
+            scale (float): scale the input image
+        """
+        self.img = img
+        self.scale = scale
+        self.width, self.height = img.shape[1], img.shape[0]
+        self._setup_figure(img)
+
+    def _setup_figure(self, img):
+        """
+        Args:
+            Same as in :meth:`__init__()`.
+
+        Returns:
+            fig (matplotlib.pyplot.figure): top level container for all the image plot elements.
+            ax (matplotlib.pyplot.Axes): contains figure elements and sets the coordinate system.
+        """
+        fig = mplfigure.Figure(frameon=False)
+        self.dpi = fig.get_dpi()
+        # add a small 1e-2 to avoid precision lost due to matplotlib's truncation
+        # (https://github.com/matplotlib/matplotlib/issues/15363)
+        fig.set_size_inches(
+            (self.width * self.scale + 1e-2) / self.dpi,
+            (self.height * self.scale + 1e-2) / self.dpi,
+        )
+        self.canvas = FigureCanvasAgg(fig)
+        # self.canvas = mpl.backends.backend_cairo.FigureCanvasCairo(fig)
+        ax = fig.add_axes([0.0, 0.0, 1.0, 1.0])
+        ax.axis("off")
+        self.fig = fig
+        self.ax = ax
+        self.reset_image(img)
+
+    def reset_image(self, img):
+        """
+        Args:
+            img: same as in __init__
+        """
+        img = img.astype("uint8")
+        self.ax.imshow(img, extent=(0, self.width, self.height, 0), interpolation="nearest")
+
+    def save(self, filepath):
+        """
+        Args:
+            filepath (str): a string that contains the absolute path, including the file name, where
+                the visualized image will be saved.
+        """
+        self.fig.savefig(filepath)
+
+    def get_image(self):
+        """
+        Returns:
+            ndarray:
+                the visualized image of shape (H, W, 3) (RGB) in uint8 type.
+                The shape is scaled w.r.t the input image using the given `scale` argument.
+        """
+        canvas = self.canvas
+        s, (width, height) = canvas.print_to_buffer()
+        # buf = io.BytesIO()  # works for cairo backend
+        # canvas.print_rgba(buf)
+        # width, height = self.width, self.height
+        # s = buf.getvalue()
+
+        buffer = np.frombuffer(s, dtype="uint8")
+
+        img_rgba = buffer.reshape(height, width, 4)
+        rgb, alpha = np.split(img_rgba, [3], axis=2)
+        return rgb.astype("uint8")
+
+
+class Visualizer:
+    """Visualizer that draws data about detection/segmentation on images.
+
+    It contains methods like `draw_{text,box,circle,line,binary_mask,polygon}`
+    that draw primitive objects to images, as well as high-level wrappers like
+    `draw_{instance_predictions,sem_seg,panoptic_seg_predictions,dataset_dict}`
+    that draw composite data in some pre-defined style.
+
+    Note that the exact visualization style for the high-level wrappers are subject to change.
+    Style such as color, opacity, label contents, visibility of labels, or even the visibility
+    of objects themselves (e.g. when the object is too small) may change according
+    to different heuristics, as long as the results still look visually reasonable.
+
+    To obtain a consistent style, you can implement custom drawing functions with the
+    abovementioned primitive methods instead. If you need more customized visualization
+    styles, you can process the data yourself following their format documented in
+    tutorials (:doc:`/tutorials/models`, :doc:`/tutorials/datasets`). This class does not
+    intend to satisfy everyone's preference on drawing styles.
+
+    This visualizer focuses on high rendering quality rather than performance. It is not
+    designed to be used for real-time applications.
+    """
+
+    # TODO implement a fast, rasterized version using OpenCV
+
+    def __init__(self, img_rgb, metadata=None, scale=1.0, instance_mode=ColorMode.IMAGE):
+        """
+        Args:
+            img_rgb: a numpy array of shape (H, W, C), where H and W correspond to
+                the height and width of the image respectively. C is the number of
+                color channels. The image is required to be in RGB format since that
+                is a requirement of the Matplotlib library. The image is also expected
+                to be in the range [0, 255].
+            metadata (Metadata): dataset metadata (e.g. class names and colors)
+            instance_mode (ColorMode): defines one of the pre-defined style for drawing
+                instances on an image.
+        """
+        self.img = np.asarray(img_rgb).clip(0, 255).astype(np.uint8)
+        if metadata is None:
+            metadata = MetadataCatalog.get("__nonexist__")
+        self.metadata = metadata
+        self.output = VisImage(self.img, scale=scale)
+        self.cpu_device = torch.device("cpu")
+
+        # too small texts are useless, therefore clamp to 9
+        self._default_font_size = max(np.sqrt(self.output.height * self.output.width) // 90, 10 // scale)
+        self._default_font_size = 18
+        self._instance_mode = instance_mode
+        self.keypoint_threshold = _KEYPOINT_THRESHOLD
+
+        import matplotlib.colors as mcolors
+
+        css4_colors = mcolors.CSS4_COLORS
+        self.color_proposals = [list(mcolors.hex2color(color)) for color in css4_colors.values()]
+
+    def draw_instance_predictions(self, predictions):
+        """Draw instance-level prediction results on an image.
+
+        Args:
+            predictions (Instances): the output of an instance detection/segmentation
+                model. Following fields will be used to draw:
+                "pred_boxes", "pred_classes", "scores", "pred_masks" (or "pred_masks_rle").
+
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        boxes = predictions.pred_boxes if predictions.has("pred_boxes") else None
+        scores = predictions.scores if predictions.has("scores") else None
+        classes = predictions.pred_classes.tolist() if predictions.has("pred_classes") else None
+        labels = _create_text_labels(classes, scores, self.metadata.get("thing_classes", None))
+        keypoints = predictions.pred_keypoints if predictions.has("pred_keypoints") else None
+
+        keep = (scores > 0.5).cpu()
+        boxes = boxes[keep]
+        scores = scores[keep]
+        classes = np.array(classes)
+        classes = classes[np.array(keep)]
+        labels = np.array(labels)
+        labels = labels[np.array(keep)]
+
+        if predictions.has("pred_masks"):
+            masks = np.asarray(predictions.pred_masks)
+            masks = masks[np.array(keep)]
+            masks = [GenericMask(x, self.output.height, self.output.width) for x in masks]
+        else:
+            masks = None
+
+        if self._instance_mode == ColorMode.SEGMENTATION and self.metadata.get("thing_colors"):
+            # if self.metadata.get("thing_colors"):
+            colors = [self._jitter([x / 255 for x in self.metadata.thing_colors[c]]) for c in classes]
+            alpha = 0.4
+        else:
+            colors = None
+            alpha = 0.4
+
+        if self._instance_mode == ColorMode.IMAGE_BW:
+            self.output.reset_image(
+                self._create_grayscale_image(
+                    (predictions.pred_masks.any(dim=0) > 0).numpy() if predictions.has("pred_masks") else None
+                )
+            )
+            alpha = 0.3
+
+        self.overlay_instances(
+            masks=masks,
+            boxes=boxes,
+            labels=labels,
+            keypoints=keypoints,
+            assigned_colors=colors,
+            alpha=alpha,
+        )
+        return self.output
+
+    def draw_sem_seg(self, sem_seg, area_threshold=None, alpha=0.7):
+        """Draw semantic segmentation predictions/labels.
+
+        Args:
+            sem_seg (Tensor or ndarray): the segmentation of shape (H, W).
+                Each value is the integer label of the pixel.
+            area_threshold (int): segments with less than `area_threshold` are not drawn.
+            alpha (float): the larger it is, the more opaque the segmentations are.
+
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        if isinstance(sem_seg, torch.Tensor):
+            sem_seg = sem_seg.numpy()
+        labels, areas = np.unique(sem_seg, return_counts=True)
+        sorted_idxs = np.argsort(-areas).tolist()
+        labels = labels[sorted_idxs]
+        for label in filter(lambda l: l < len(self.metadata.stuff_classes), labels):
+            try:
+                mask_color = [x / 255 for x in self.metadata.stuff_colors[label]]
+            except (AttributeError, IndexError):
+                mask_color = None
+
+            binary_mask = (sem_seg == label).astype(np.uint8)
+            text = self.metadata.stuff_classes[label]
+            self.draw_binary_mask(
+                binary_mask,
+                color=mask_color,
+                edge_color=_OFF_WHITE,
+                text=text,
+                alpha=alpha,
+                area_threshold=area_threshold,
+            )
+        return self.output
+
+    def draw_panoptic_seg(self, panoptic_seg, segments_info, area_threshold=None, alpha=0.7):
+        """Draw panoptic prediction annotations or results.
+
+        Args:
+            panoptic_seg (Tensor): of shape (height, width) where the values are ids for each
+                segment.
+            segments_info (list[dict] or None): Describe each segment in `panoptic_seg`.
+                If it is a ``list[dict]``, each dict contains keys "id", "category_id".
+                If None, category id of each pixel is computed by
+                ``pixel // metadata.label_divisor``.
+            area_threshold (int): stuff segments with less than `area_threshold` are not drawn.
+
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        pred = _PanopticPrediction(panoptic_seg, segments_info, self.metadata)
+
+        if self._instance_mode == ColorMode.IMAGE_BW:
+            self.output.reset_image(self._create_grayscale_image(pred.non_empty_mask()))
+
+        # draw mask for all semantic segments first i.e. "stuff"
+        for mask, sinfo in pred.semantic_masks():
+            category_idx = sinfo["category_id"]
+            try:
+                mask_color = [x / 255 for x in self.metadata.stuff_colors[category_idx]]
+            except AttributeError:
+                mask_color = None
+
+            text = self.metadata.stuff_classes[category_idx].replace("-other", "").replace("-merged", "")
+            self.draw_binary_mask(
+                mask,
+                color=mask_color,
+                edge_color=_OFF_WHITE,
+                text=text,
+                alpha=alpha,
+                area_threshold=area_threshold,
+            )
+
+        # draw mask for all instances second
+        all_instances = list(pred.instance_masks())
+        if len(all_instances) == 0:
+            return self.output
+        masks, sinfo = list(zip(*all_instances))
+        category_ids = [x["category_id"] for x in sinfo]
+
+        try:
+            scores = [x["score"] for x in sinfo]
+        except KeyError:
+            scores = None
+        class_names = [name.replace("-other", "").replace("-merged", "") for name in self.metadata.thing_classes]
+        labels = _create_text_labels(category_ids, scores, class_names, [x.get("iscrowd", 0) for x in sinfo])
+
+        try:
+            colors = [self._jitter([x / 255 for x in self.metadata.thing_colors[c]]) for c in category_ids]
+        except AttributeError:
+            colors = None
+        self.overlay_instances(masks=masks, labels=labels, assigned_colors=colors, alpha=alpha)
+
+        return self.output
+
+    draw_panoptic_seg_predictions = draw_panoptic_seg  # backward compatibility
+
+    def draw_dataset_dict(self, dic):
+        """Draw annotations/segmentaions in Detectron2 Dataset format.
+
+        Args:
+            dic (dict): annotation/segmentation data of one image, in Detectron2 Dataset format.
+
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        annos = dic.get("annotations", None)
+        if annos:
+            if "segmentation" in annos[0]:
+                masks = [x["segmentation"] for x in annos]
+            else:
+                masks = None
+            if "keypoints" in annos[0]:
+                keypts = [x["keypoints"] for x in annos]
+                keypts = np.array(keypts).reshape(len(annos), -1, 3)
+            else:
+                keypts = None
+
+            boxes = [
+                BoxMode.convert(x["bbox"], x["bbox_mode"], BoxMode.XYXY_ABS) if len(x["bbox"]) == 4 else x["bbox"]
+                for x in annos
+            ]
+
+            colors = None
+            category_ids = [x["category_id"] for x in annos]
+            if self._instance_mode == ColorMode.SEGMENTATION and self.metadata.get("thing_colors"):
+                colors = [self._jitter([x / 255 for x in self.metadata.thing_colors[c]]) for c in category_ids]
+            names = self.metadata.get("thing_classes", None)
+            labels = _create_text_labels(
+                category_ids,
+                scores=None,
+                class_names=names,
+                is_crowd=[x.get("iscrowd", 0) for x in annos],
+            )
+            self.overlay_instances(
+                labels=labels,
+                boxes=boxes,
+                masks=masks,
+                keypoints=keypts,
+                assigned_colors=colors,
+            )
+
+        sem_seg = dic.get("sem_seg", None)
+        if sem_seg is None and "sem_seg_file_name" in dic:
+            with PathManager.open(dic["sem_seg_file_name"], "rb") as f:
+                sem_seg = Image.open(f)
+                sem_seg = np.asarray(sem_seg, dtype="uint8")
+        if sem_seg is not None:
+            self.draw_sem_seg(sem_seg, area_threshold=0, alpha=0.4)
+
+        pan_seg = dic.get("pan_seg", None)
+        if pan_seg is None and "pan_seg_file_name" in dic:
+            with PathManager.open(dic["pan_seg_file_name"], "rb") as f:
+                pan_seg = Image.open(f)
+                pan_seg = np.asarray(pan_seg)
+                from panopticapi.utils import rgb2id
+
+                pan_seg = rgb2id(pan_seg)
+        if pan_seg is not None:
+            segments_info = dic["segments_info"]
+            pan_seg = torch.tensor(pan_seg)
+            self.draw_panoptic_seg(pan_seg, segments_info, area_threshold=0, alpha=0.7)
+        return self.output
+
+    def overlay_instances(
+        self,
+        *,
+        boxes=None,
+        labels=None,
+        masks=None,
+        keypoints=None,
+        assigned_colors=None,
+        alpha=0.5,
+    ):
+        """
+        Args:
+            boxes (Boxes, RotatedBoxes or ndarray): either a :class:`Boxes`,
+                or an Nx4 numpy array of XYXY_ABS format for the N objects in a single image,
+                or a :class:`RotatedBoxes`,
+                or an Nx5 numpy array of (x_center, y_center, width, height, angle_degrees) format
+                for the N objects in a single image,
+            labels (list[str]): the text to be displayed for each instance.
+            masks (masks-like object): Supported types are:
+
+                * :class:`detectron2.structures.PolygonMasks`,
+                  :class:`detectron2.structures.BitMasks`.
+                * list[list[ndarray]]: contains the segmentation masks for all objects in one image.
+                  The first level of the list corresponds to individual instances. The second
+                  level to all the polygon that compose the instance, and the third level
+                  to the polygon coordinates. The third level should have the format of
+                  [x0, y0, x1, y1, ..., xn, yn] (n >= 3).
+                * list[ndarray]: each ndarray is a binary mask of shape (H, W).
+                * list[dict]: each dict is a COCO-style RLE.
+            keypoints (Keypoint or array like): an array-like object of shape (N, K, 3),
+                where the N is the number of instances and K is the number of keypoints.
+                The last dimension corresponds to (x, y, visibility or score).
+            assigned_colors (list[matplotlib.colors]): a list of colors, where each color
+                corresponds to each mask or box in the image. Refer to 'matplotlib.colors'
+                for full list of formats that the colors are accepted in.
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        num_instances = 0
+        if boxes is not None:
+            boxes = self._convert_boxes(boxes)
+            num_instances = len(boxes)
+        if masks is not None:
+            masks = self._convert_masks(masks)
+            if num_instances:
+                assert len(masks) == num_instances
+            else:
+                num_instances = len(masks)
+        if keypoints is not None:
+            if num_instances:
+                assert len(keypoints) == num_instances
+            else:
+                num_instances = len(keypoints)
+            keypoints = self._convert_keypoints(keypoints)
+        if labels is not None:
+            assert len(labels) == num_instances
+        if assigned_colors is None:
+            assigned_colors = [random_color(rgb=True, maximum=1) for _ in range(num_instances)]
+        if num_instances == 0:
+            return self.output
+        if boxes is not None and boxes.shape[1] == 5:
+            return self.overlay_rotated_instances(boxes=boxes, labels=labels, assigned_colors=assigned_colors)
+
+        # Display in largest to smallest order to reduce occlusion.
+        areas = None
+        if boxes is not None:
+            areas = np.prod(boxes[:, 2:] - boxes[:, :2], axis=1)
+        elif masks is not None:
+            areas = np.asarray([x.area() for x in masks])
+
+        if areas is not None:
+            sorted_idxs = np.argsort(-areas).tolist()
+            # Re-order overlapped instances in descending order.
+            boxes = boxes[sorted_idxs] if boxes is not None else None
+            labels = [labels[k] for k in sorted_idxs] if labels is not None else None
+            masks = [masks[idx] for idx in sorted_idxs] if masks is not None else None
+            assigned_colors = [assigned_colors[idx] for idx in sorted_idxs]
+            keypoints = keypoints[sorted_idxs] if keypoints is not None else None
+
+        for i in range(num_instances):
+            color = assigned_colors[i]
+            if boxes is not None:
+                self.draw_box(boxes[i], edge_color=color)
+
+            if masks is not None:
+                for segment in masks[i].polygons:
+                    self.draw_polygon(segment.reshape(-1, 2), color, alpha=alpha)
+
+            if labels is not None:
+                # first get a box
+                if boxes is not None:
+                    x0, y0, x1, y1 = boxes[i]
+                    text_pos = (x0, y0)  # if drawing boxes, put text on the box corner.
+                    horiz_align = "left"
+                elif masks is not None:
+                    # skip small mask without polygon
+                    if len(masks[i].polygons) == 0:
+                        continue
+
+                    x0, y0, x1, y1 = masks[i].bbox()
+
+                    # draw text in the center (defined by median) when box is not drawn
+                    # median is less sensitive to outliers.
+                    text_pos = np.median(masks[i].mask.nonzero(), axis=1)[::-1]
+                    horiz_align = "center"
+                else:
+                    continue  # drawing the box confidence for keypoints isn't very useful.
+                # for small objects, draw text at the side to avoid occlusion
+                instance_area = (y1 - y0) * (x1 - x0)
+                if instance_area < _SMALL_OBJECT_AREA_THRESH * self.output.scale or y1 - y0 < 40 * self.output.scale:
+                    if y1 >= self.output.height - 5:
+                        text_pos = (x1, y0)
+                    else:
+                        text_pos = (x0, y1)
+
+                height_ratio = (y1 - y0) / np.sqrt(self.output.height * self.output.width)
+                lighter_color = self._change_color_brightness(color, brightness_factor=0.7)
+                font_size = np.clip((height_ratio - 0.02) / 0.08 + 1, 1.2, 2) * 0.5 * self._default_font_size
+                self.draw_text(
+                    labels[i],
+                    text_pos,
+                    color=lighter_color,
+                    horizontal_alignment=horiz_align,
+                    font_size=font_size,
+                )
+
+        # draw keypoints
+        if keypoints is not None:
+            for keypoints_per_instance in keypoints:
+                self.draw_and_connect_keypoints(keypoints_per_instance)
+
+        return self.output
+
+    def overlay_rotated_instances(self, boxes=None, labels=None, assigned_colors=None):
+        """
+        Args:
+            boxes (ndarray): an Nx5 numpy array of
+                (x_center, y_center, width, height, angle_degrees) format
+                for the N objects in a single image.
+            labels (list[str]): the text to be displayed for each instance.
+            assigned_colors (list[matplotlib.colors]): a list of colors, where each color
+                corresponds to each mask or box in the image. Refer to 'matplotlib.colors'
+                for full list of formats that the colors are accepted in.
+
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        num_instances = len(boxes)
+
+        if assigned_colors is None:
+            assigned_colors = [random_color(rgb=True, maximum=1) for _ in range(num_instances)]
+        if num_instances == 0:
+            return self.output
+
+        # Display in largest to smallest order to reduce occlusion.
+        if boxes is not None:
+            areas = boxes[:, 2] * boxes[:, 3]
+
+        sorted_idxs = np.argsort(-areas).tolist()
+        # Re-order overlapped instances in descending order.
+        boxes = boxes[sorted_idxs]
+        labels = [labels[k] for k in sorted_idxs] if labels is not None else None
+        colors = [assigned_colors[idx] for idx in sorted_idxs]
+
+        for i in range(num_instances):
+            self.draw_rotated_box_with_label(
+                boxes[i],
+                edge_color=colors[i],
+                label=labels[i] if labels is not None else None,
+            )
+
+        return self.output
+
+    def draw_and_connect_keypoints(self, keypoints):
+        """Draws keypoints of an instance and follows the rules for keypoint connections to draw lines between
+        appropriate keypoints. This follows color heuristics for line color.
+
+        Args:
+            keypoints (Tensor): a tensor of shape (K, 3), where K is the number of keypoints
+                and the last dimension corresponds to (x, y, probability).
+
+        Returns:
+            output (VisImage): image object with visualizations.
+        """
+        visible = {}
+        keypoint_names = self.metadata.get("keypoint_names")
+        for idx, keypoint in enumerate(keypoints):
+            # draw keypoint
+            x, y, prob = keypoint
+            if prob > self.keypoint_threshold:
+                self.draw_circle((x, y), color=_RED)
+                if keypoint_names:
+                    keypoint_name = keypoint_names[idx]
+                    visible[keypoint_name] = (x, y)
+
+        if self.metadata.get("keypoint_connection_rules"):
+            for kp0, kp1, color in self.metadata.keypoint_connection_rules:
+                if kp0 in visible and kp1 in visible:
+                    x0, y0 = visible[kp0]
+                    x1, y1 = visible[kp1]
+                    color = tuple(x / 255.0 for x in color)
+                    self.draw_line([x0, x1], [y0, y1], color=color)
+
+        # draw lines from nose to mid-shoulder and mid-shoulder to mid-hip
+        # Note that this strategy is specific to person keypoints.
+        # For other keypoints, it should just do nothing
+        try:
+            ls_x, ls_y = visible["left_shoulder"]
+            rs_x, rs_y = visible["right_shoulder"]
+            mid_shoulder_x, mid_shoulder_y = (ls_x + rs_x) / 2, (ls_y + rs_y) / 2
+        except KeyError:
+            pass
+        else:
+            # draw line from nose to mid-shoulder
+            nose_x, nose_y = visible.get("nose", (None, None))
+            if nose_x is not None:
+                self.draw_line([nose_x, mid_shoulder_x], [nose_y, mid_shoulder_y], color=_RED)
+
+            try:
+                # draw line from mid-shoulder to mid-hip
+                lh_x, lh_y = visible["left_hip"]
+                rh_x, rh_y = visible["right_hip"]
+            except KeyError:
+                pass
+            else:
+                mid_hip_x, mid_hip_y = (lh_x + rh_x) / 2, (lh_y + rh_y) / 2
+                self.draw_line([mid_hip_x, mid_shoulder_x], [mid_hip_y, mid_shoulder_y], color=_RED)
+        return self.output
+
+    """
+    Primitive drawing functions:
+    """
+
+    def draw_text(
+        self,
+        text,
+        position,
+        *,
+        font_size=None,
+        color="g",
+        horizontal_alignment="center",
+        rotation=0,
+    ):
+        """
+        Args:
+            text (str): class label
+            position (tuple): a tuple of the x and y coordinates to place text on image.
+            font_size (int, optional): font of the text. If not provided, a font size
+                proportional to the image width is calculated and used.
+            color: color of the text. Refer to `matplotlib.colors` for full list
+                of formats that are accepted.
+            horizontal_alignment (str): see `matplotlib.text.Text`
+            rotation: rotation angle in degrees CCW
+
+        Returns:
+            output (VisImage): image object with text drawn.
+        """
+        if not font_size:
+            font_size = self._default_font_size
+
+        # since the text background is dark, we don't want the text to be dark
+        color = np.maximum(list(mplc.to_rgb(color)), 0.15)
+        color[np.argmax(color)] = max(0.8, np.max(color))
+
+        def contrasting_color(rgb):
+            """Returns 'white' or 'black' depending on which color contrasts more with the given RGB value."""
+
+            # Decompose the RGB tuple
+            R, G, B = rgb
+
+            # Calculate the Y value
+            Y = 0.299 * R + 0.587 * G + 0.114 * B
+
+            # If Y value is greater than 128, it's closer to white so return black. Otherwise, return white.
+            return "black" if Y > 128 else "white"
+
+        bbox_background = contrasting_color(color * 255)
+
+        x, y = position
+        self.output.ax.text(
+            x,
+            y,
+            text,
+            size=font_size * self.output.scale,
+            family="sans-serif",
+            bbox={
+                "facecolor": bbox_background,
+                "alpha": 0.8,
+                "pad": 0.7,
+                "edgecolor": "none",
+            },
+            verticalalignment="top",
+            horizontalalignment=horizontal_alignment,
+            color=color,
+            zorder=10,
+            rotation=rotation,
+        )
+        return self.output
+
+    def draw_box(self, box_coord, alpha=0.5, edge_color="g", line_style="-"):
+        """
+        Args:
+            box_coord (tuple): a tuple containing x0, y0, x1, y1 coordinates, where x0 and y0
+                are the coordinates of the image's top left corner. x1 and y1 are the
+                coordinates of the image's bottom right corner.
+            alpha (float): blending efficient. Smaller values lead to more transparent masks.
+            edge_color: color of the outline of the box. Refer to `matplotlib.colors`
+                for full list of formats that are accepted.
+            line_style (string): the string to use to create the outline of the boxes.
+
+        Returns:
+            output (VisImage): image object with box drawn.
+        """
+        x0, y0, x1, y1 = box_coord
+        width = x1 - x0
+        height = y1 - y0
+
+        linewidth = max(self._default_font_size / 12, 1)
+
+        self.output.ax.add_patch(
+            mpl.patches.Rectangle(
+                (x0, y0),
+                width,
+                height,
+                fill=False,
+                edgecolor=edge_color,
+                linewidth=linewidth * self.output.scale,
+                alpha=alpha,
+                linestyle=line_style,
+            )
+        )
+        return self.output
+
+    def draw_rotated_box_with_label(self, rotated_box, alpha=0.5, edge_color="g", line_style="-", label=None):
+        """Draw a rotated box with label on its top-left corner.
+
+        Args:
+            rotated_box (tuple): a tuple containing (cnt_x, cnt_y, w, h, angle),
+                where cnt_x and cnt_y are the center coordinates of the box.
+                w and h are the width and height of the box. angle represents how
+                many degrees the box is rotated CCW with regard to the 0-degree box.
+            alpha (float): blending efficient. Smaller values lead to more transparent masks.
+            edge_color: color of the outline of the box. Refer to `matplotlib.colors`
+                for full list of formats that are accepted.
+            line_style (string): the string to use to create the outline of the boxes.
+            label (string): label for rotated box. It will not be rendered when set to None.
+
+        Returns:
+            output (VisImage): image object with box drawn.
+        """
+        cnt_x, cnt_y, w, h, angle = rotated_box
+        area = w * h
+        # use thinner lines when the box is small
+        linewidth = self._default_font_size / (6 if area < _SMALL_OBJECT_AREA_THRESH * self.output.scale else 3)
+
+        theta = angle * math.pi / 180.0
+        c = math.cos(theta)
+        s = math.sin(theta)
+        rect = [(-w / 2, h / 2), (-w / 2, -h / 2), (w / 2, -h / 2), (w / 2, h / 2)]
+        # x: left->right ; y: top->down
+        rotated_rect = [(s * yy + c * xx + cnt_x, c * yy - s * xx + cnt_y) for (xx, yy) in rect]
+        for k in range(4):
+            j = (k + 1) % 4
+            self.draw_line(
+                [rotated_rect[k][0], rotated_rect[j][0]],
+                [rotated_rect[k][1], rotated_rect[j][1]],
+                color=edge_color,
+                linestyle="--" if k == 1 else line_style,
+                linewidth=linewidth,
+            )
+
+        if label is not None:
+            text_pos = rotated_rect[1]  # topleft corner
+
+            height_ratio = h / np.sqrt(self.output.height * self.output.width)
+            label_color = self._change_color_brightness(edge_color, brightness_factor=0.7)
+            font_size = np.clip((height_ratio - 0.02) / 0.08 + 1, 1.2, 2) * 0.5 * self._default_font_size
+            self.draw_text(label, text_pos, color=label_color, font_size=font_size, rotation=angle)
+
+        return self.output
+
+    def draw_circle(self, circle_coord, color, radius=3):
+        """
+        Args:
+            circle_coord (list(int) or tuple(int)): contains the x and y coordinates
+                of the center of the circle.
+            color: color of the polygon. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted.
+            radius (int): radius of the circle.
+
+        Returns:
+            output (VisImage): image object with box drawn.
+        """
+        x, y = circle_coord
+        self.output.ax.add_patch(mpl.patches.Circle(circle_coord, radius=radius, fill=True, color=color))
+        return self.output
+
+    def draw_line(self, x_data, y_data, color, linestyle="-", linewidth=None):
+        """
+        Args:
+            x_data (list[int]): a list containing x values of all the points being drawn.
+                Length of list should match the length of y_data.
+            y_data (list[int]): a list containing y values of all the points being drawn.
+                Length of list should match the length of x_data.
+            color: color of the line. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted.
+            linestyle: style of the line. Refer to `matplotlib.lines.Line2D`
+                for a full list of formats that are accepted.
+            linewidth (float or None): width of the line. When it's None,
+                a default value will be computed and used.
+
+        Returns:
+            output (VisImage): image object with line drawn.
+        """
+        if linewidth is None:
+            linewidth = self._default_font_size / 3
+        linewidth = max(linewidth, 1)
+        self.output.ax.add_line(
+            mpl.lines.Line2D(
+                x_data,
+                y_data,
+                linewidth=linewidth * self.output.scale,
+                color=color,
+                linestyle=linestyle,
+            )
+        )
+        return self.output
+
+    def draw_binary_mask(
+        self,
+        binary_mask,
+        color=None,
+        *,
+        edge_color=None,
+        text=None,
+        alpha=0.7,
+        area_threshold=10,
+    ):
+        """
+        Args:
+            binary_mask (ndarray): numpy array of shape (H, W), where H is the image height and
+                W is the image width. Each value in the array is either a 0 or 1 value of uint8
+                type.
+            color: color of the mask. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted. If None, will pick a random color.
+            edge_color: color of the polygon edges. Refer to `matplotlib.colors` for a
+                full list of formats that are accepted.
+            text (str): if None, will be drawn on the object
+            alpha (float): blending efficient. Smaller values lead to more transparent masks.
+            area_threshold (float): a connected component smaller than this area will not be shown.
+
+        Returns:
+            output (VisImage): image object with mask drawn.
+        """
+        if color is None:
+            color = random_color(rgb=True, maximum=1)
+        color = mplc.to_rgb(color)
+
+        has_valid_segment = False
+        binary_mask = binary_mask.astype("uint8")  # opencv needs uint8
+        mask = GenericMask(binary_mask, self.output.height, self.output.width)
+        shape2d = (binary_mask.shape[0], binary_mask.shape[1])
+
+        if not mask.has_holes:
+            # draw polygons for regular masks
+            for segment in mask.polygons:
+                area = mask_util.area(mask_util.frPyObjects([segment], shape2d[0], shape2d[1]))
+                if area < (area_threshold or 0):
+                    continue
+                has_valid_segment = True
+                segment = segment.reshape(-1, 2)
+                self.draw_polygon(segment, color=color, edge_color=edge_color, alpha=alpha)
+        else:
+            # TODO: Use Path/PathPatch to draw vector graphics:
+            # https://stackoverflow.com/questions/8919719/how-to-plot-a-complex-polygon
+            rgba = np.zeros(shape2d + (4,), dtype="float32")
+            rgba[:, :, :3] = color
+            rgba[:, :, 3] = (mask.mask == 1).astype("float32") * alpha
+            has_valid_segment = True
+            self.output.ax.imshow(rgba, extent=(0, self.output.width, self.output.height, 0))
+
+        if text is not None and has_valid_segment:
+            lighter_color = self._change_color_brightness(color, brightness_factor=0.7)
+            self._draw_text_in_mask(binary_mask, text, lighter_color)
+        return self.output
+
+    def draw_binary_mask_with_number(
+        self,
+        binary_mask,
+        color=None,
+        *,
+        edge_color=None,
+        text=None,
+        label_mode="1",
+        alpha=0.1,
+        anno_mode=["Mask"],
+        area_threshold=10,
+    ):
+        """
+        Args:
+            binary_mask (ndarray): numpy array of shape (H, W), where H is the image height and
+                W is the image width. Each value in the array is either a 0 or 1 value of uint8
+                type.
+            color: color of the mask. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted. If None, will pick a random color.
+            edge_color: color of the polygon edges. Refer to `matplotlib.colors` for a
+                full list of formats that are accepted.
+            text (str): if None, will be drawn on the object
+            alpha (float): blending efficient. Smaller values lead to more transparent masks.
+            area_threshold (float): a connected component smaller than this area will not be shown.
+
+        Returns:
+            output (VisImage): image object with mask drawn.
+        """
+        if color is None:
+            randint = random.randint(0, len(self.color_proposals) - 1)
+            color = self.color_proposals[randint]
+        color = mplc.to_rgb(color)
+
+        has_valid_segment = True
+        binary_mask = binary_mask.astype("uint8")  # opencv needs uint8
+        mask = GenericMask(binary_mask, self.output.height, self.output.width)
+        shape2d = (binary_mask.shape[0], binary_mask.shape[1])
+        bbox = mask.bbox()
+
+        if "Mask" in anno_mode:
+            if not mask.has_holes:
+                # draw polygons for regular masks
+                for segment in mask.polygons:
+                    area = mask_util.area(mask_util.frPyObjects([segment], shape2d[0], shape2d[1]))
+                    if area < (area_threshold or 0):
+                        continue
+                    has_valid_segment = True
+                    segment = segment.reshape(-1, 2)
+                    self.draw_polygon(segment, color=color, edge_color=edge_color, alpha=alpha)
+            else:
+                # TODO: Use Path/PathPatch to draw vector graphics:
+                # https://stackoverflow.com/questions/8919719/how-to-plot-a-complex-polygon
+                rgba = np.zeros(shape2d + (4,), dtype="float32")
+                rgba[:, :, :3] = color
+                rgba[:, :, 3] = (mask.mask == 1).astype("float32") * alpha
+                has_valid_segment = True
+                self.output.ax.imshow(rgba, extent=(0, self.output.width, self.output.height, 0))
+
+        if "Box" in anno_mode:
+            self.draw_box(bbox, edge_color=color, alpha=0.75)
+
+        if "Mark" in anno_mode:
+            has_valid_segment = True
+        else:
+            has_valid_segment = False
+
+        if text is not None and has_valid_segment:
+            # lighter_color = tuple([x*0.2 for x in color])
+            lighter_color = [
+                1,
+                1,
+                1,
+            ]  # self._change_color_brightness(color, brightness_factor=0.7)
+            self._draw_number_in_mask(binary_mask, text, lighter_color, label_mode)
+        return self.output
+
+    def draw_soft_mask(self, soft_mask, color=None, *, text=None, alpha=0.5):
+        """
+        Args:
+            soft_mask (ndarray): float array of shape (H, W), each value in [0, 1].
+            color: color of the mask. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted. If None, will pick a random color.
+            text (str): if None, will be drawn on the object
+            alpha (float): blending efficient. Smaller values lead to more transparent masks.
+
+        Returns:
+            output (VisImage): image object with mask drawn.
+        """
+        if color is None:
+            color = random_color(rgb=True, maximum=1)
+        color = mplc.to_rgb(color)
+
+        shape2d = (soft_mask.shape[0], soft_mask.shape[1])
+        rgba = np.zeros(shape2d + (4,), dtype="float32")
+        rgba[:, :, :3] = color
+        rgba[:, :, 3] = soft_mask * alpha
+        self.output.ax.imshow(rgba, extent=(0, self.output.width, self.output.height, 0))
+
+        if text is not None:
+            lighter_color = self._change_color_brightness(color, brightness_factor=0.7)
+            binary_mask = (soft_mask > 0.5).astype("uint8")
+            self._draw_text_in_mask(binary_mask, text, lighter_color)
+        return self.output
+
+    def draw_polygon(self, segment, color, edge_color=None, alpha=0.5):
+        """
+        Args:
+            segment: numpy array of shape Nx2, containing all the points in the polygon.
+            color: color of the polygon. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted.
+            edge_color: color of the polygon edges. Refer to `matplotlib.colors` for a
+                full list of formats that are accepted. If not provided, a darker shade
+                of the polygon color will be used instead.
+            alpha (float): blending efficient. Smaller values lead to more transparent masks.
+
+        Returns:
+            output (VisImage): image object with polygon drawn.
+        """
+        if edge_color is None:
+            # make edge color darker than the polygon color
+            if alpha > 0.8:
+                edge_color = self._change_color_brightness(color, brightness_factor=-0.7)
+            else:
+                edge_color = color
+        edge_color = mplc.to_rgb(edge_color) + (1,)
+
+        polygon = mpl.patches.Polygon(
+            segment,
+            fill=True,
+            facecolor=mplc.to_rgb(color) + (alpha,),
+            edgecolor=edge_color,
+            linewidth=max(self._default_font_size // 15 * self.output.scale, 1),
+        )
+        self.output.ax.add_patch(polygon)
+        return self.output
+
+    """
+    Internal methods:
+    """
+
+    def _jitter(self, color):
+        """Randomly modifies given color to produce a slightly different color than the color given.
+
+        Args:
+            color (tuple[double]): a tuple of 3 elements, containing the RGB values of the color
+                picked. The values in the list are in the [0.0, 1.0] range.
+
+        Returns:
+            jittered_color (tuple[double]): a tuple of 3 elements, containing the RGB values of the
+                color after being jittered. The values in the list are in the [0.0, 1.0] range.
+        """
+        color = mplc.to_rgb(color)
+        # np.random.seed(0)
+        vec = np.random.rand(3)
+        # better to do it in another color space
+        vec = vec / np.linalg.norm(vec) * 0.5
+        res = np.clip(vec + color, 0, 1)
+        return tuple(res)
+
+    def _create_grayscale_image(self, mask=None):
+        """Create a grayscale version of the original image.
+
+        The colors in masked area, if given, will be kept.
+        """
+        img_bw = self.img.astype("f4").mean(axis=2)
+        img_bw = np.stack([img_bw] * 3, axis=2)
+        if mask is not None:
+            img_bw[mask] = self.img[mask]
+        return img_bw
+
+    def _change_color_brightness(self, color, brightness_factor):
+        """Depending on the brightness_factor, gives a lighter or darker color i.e. a color with less or more saturation
+        than the original color.
+
+        Args:
+            color: color of the polygon. Refer to `matplotlib.colors` for a full list of
+                formats that are accepted.
+            brightness_factor (float): a value in [-1.0, 1.0] range. A lightness factor of
+                0 will correspond to no change, a factor in [-1.0, 0) range will result in
+                a darker color and a factor in (0, 1.0] range will result in a lighter color.
+
+        Returns:
+            modified_color (tuple[double]): a tuple containing the RGB values of the
+                modified color. Each value in the tuple is in the [0.0, 1.0] range.
+        """
+        assert brightness_factor >= -1.0 and brightness_factor <= 1.0
+        color = mplc.to_rgb(color)
+        polygon_color = colorsys.rgb_to_hls(*mplc.to_rgb(color))
+        modified_lightness = polygon_color[1] + (brightness_factor * polygon_color[1])
+        modified_lightness = 0.0 if modified_lightness < 0.0 else modified_lightness
+        modified_lightness = 1.0 if modified_lightness > 1.0 else modified_lightness
+        modified_color = colorsys.hls_to_rgb(polygon_color[0], modified_lightness, polygon_color[2])
+        return modified_color
+
+    def _convert_boxes(self, boxes):
+        """Convert different format of boxes to an NxB array, where B = 4 or 5 is the box dimension."""
+        if isinstance(boxes, Boxes) or isinstance(boxes, RotatedBoxes):
+            return boxes.tensor.detach().numpy()
+        else:
+            return np.asarray(boxes)
+
+    def _convert_masks(self, masks_or_polygons):
+        """Convert different format of masks or polygons to a tuple of masks and polygons.
+
+        Returns:
+            list[GenericMask]:
+        """
+
+        m = masks_or_polygons
+        if isinstance(m, PolygonMasks):
+            m = m.polygons
+        if isinstance(m, BitMasks):
+            m = m.tensor.numpy()
+        if isinstance(m, torch.Tensor):
+            m = m.numpy()
+        ret = []
+        for x in m:
+            if isinstance(x, GenericMask):
+                ret.append(x)
+            else:
+                ret.append(GenericMask(x, self.output.height, self.output.width))
+        return ret
+
+    def _draw_number_in_mask(self, binary_mask, text, color, label_mode="1"):
+        """Find proper places to draw text given a binary mask."""
+
+        def number_to_string(n):
+            chars = []
+            while n:
+                n, remainder = divmod(n - 1, 26)
+                chars.append(chr(97 + remainder))
+            return "".join(reversed(chars))
+
+        binary_mask = np.pad(binary_mask, ((1, 1), (1, 1)), "constant")
+        mask_dt = cv2.distanceTransform(binary_mask, cv2.DIST_L2, 0)
+        mask_dt = mask_dt[1:-1, 1:-1]
+        max_dist = np.max(mask_dt)
+        coords_y, coords_x = np.where(mask_dt == max_dist)  # coords is [y, x]
+
+        if label_mode == "a":
+            text = number_to_string(int(text))
+        else:
+            text = text
+
+        self.draw_text(
+            text,
+            (coords_x[len(coords_x) // 2] + 2, coords_y[len(coords_y) // 2] - 6),
+            color=color,
+        )
+
+        # TODO sometimes drawn on wrong objects. the heuristics here can improve.
+        # _num_cc, cc_labels, stats, centroids = cv2.connectedComponentsWithStats(binary_mask, 8)
+        # if stats[1:, -1].size == 0:
+        #     return
+        # largest_component_id = np.argmax(stats[1:, -1]) + 1
+
+        # # draw text on the largest component, as well as other very large components.
+        # for cid in range(1, _num_cc):
+        #     if cid == largest_component_id or stats[cid, -1] > _LARGE_MASK_AREA_THRESH:
+        #         # median is more stable than centroid
+        #         # center = centroids[largest_component_id]
+        #         center = np.median((cc_labels == cid).nonzero(), axis=1)[::-1]
+        #         # bottom=np.max((cc_labels == cid).nonzero(), axis=1)[::-1]
+        #         # center[1]=bottom[1]+2
+        #         self.draw_text(text, center, color=color)
+
+    def _draw_text_in_mask(self, binary_mask, text, color):
+        """Find proper places to draw text given a binary mask."""
+        # TODO sometimes drawn on wrong objects. the heuristics here can improve.
+        _num_cc, cc_labels, stats, centroids = cv2.connectedComponentsWithStats(binary_mask, 8)
+        if stats[1:, -1].size == 0:
+            return
+        largest_component_id = np.argmax(stats[1:, -1]) + 1
+
+        # draw text on the largest component, as well as other very large components.
+        for cid in range(1, _num_cc):
+            if cid == largest_component_id or stats[cid, -1] > _LARGE_MASK_AREA_THRESH:
+                # median is more stable than centroid
+                # center = centroids[largest_component_id]
+                center = np.median((cc_labels == cid).nonzero(), axis=1)[::-1]
+                bottom = np.max((cc_labels == cid).nonzero(), axis=1)[::-1]
+                center[1] = bottom[1] + 2
+                self.draw_text(text, center, color=color)
+
+    def _convert_keypoints(self, keypoints):
+        if isinstance(keypoints, Keypoints):
+            keypoints = keypoints.tensor
+        keypoints = np.asarray(keypoints)
+        return keypoints
+
+    def get_output(self):
+        """
+        Returns:
+            output (VisImage): the image output containing the visualizations added
+            to the image.
+        """
+        return self.output