Improve dataset card: add metadata, links and documentation

README.md

@@ -1,26 +1,69 @@
----
-license: apache-2.0
-configs:
-- config_name: default
-  data_files:
-  - split: train
-    path: data/train-*
-dataset_info:
-  features:
-  - name: camera_images
-    list: image
-  - name: depth_images
-    list: image
-  - name: normal_images
-    list: image
-  - name: frame_id
-    dtype: int32
-  - name: scene_id
-    dtype: string
-  splits:
-  - name: train
-    num_bytes: 3671744232.849
-    num_examples: 1473
-  download_size: 3336228908
-  dataset_size: 3671744232.849
----
+---
+license: apache-2.0
+task_categories:
+- robotics
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+dataset_info:
+  features:
+  - name: camera_images
+    list: image
+  - name: depth_images
+    list: image
+  - name: normal_images
+    list: image
+  - name: frame_id
+    dtype: int32
+  - name: scene_id
+    dtype: string
+  splits:
+  - name: train
+    num_bytes: 3671744232.849
+    num_examples: 1473
+  download_size: 3336228908
+  dataset_size: 3671744232.849
+---
+
+# RoboTransfer-RealData
+
+[**Project Page**](https://horizonrobotics.github.io/robot_lab/robotransfer) | [**Paper**](https://huggingface.co/papers/2505.23171) | [**GitHub**](https://github.com/HorizonRobotics/RoboTransfer)
+
+RoboTransfer-RealData is a real-world robotic manipulation dataset collected using the ALOHA-AgileX robot system. It was introduced as part of the paper **"RoboTransfer: Controllable Geometry-Consistent Video Diffusion for Manipulation Policy Transfer"**. 
+
+The dataset contains real-world trajectories used to evaluate policy transfer from synthetic data generated by RoboTransfer, a diffusion-based framework designed for geometry-consistent robotic data synthesis.
+
+## Dataset Description
+
+The dataset includes multi-modal visual data for robotic tasks:
+- `camera_images`: RGB frames captured from the robot's camera system.
+- `depth_images`: Corresponding depth maps for geometric conditioning.
+- `normal_images`: Estimated surface normal maps.
+- `frame_id`: The sequential index of the frame.
+- `scene_id`: Identifier for specific recorded scenes.
+
+## Usage
+
+As specified in the [RoboTransfer GitHub repository](https://github.com/HorizonRobotics/RoboTransfer), you can process raw RGB images from this dataset into the RoboTransfer format with geometric conditioning using the following script:
+
+```bash
+script/process_real.sh
+```
+
+## Citation
+
+If you use this dataset or the RoboTransfer framework in your research, please cite:
+
+```bibtex
+@misc{liu2025robotransfergeometryconsistentvideodiffusion,
+      title={RoboTransfer: Geometry-Consistent Video Diffusion for Robotic Visual Policy Transfer},
+      author={Liu Liu and Xiaofeng Wang and Guosheng Zhao and Keyu Li and Wenkang Qin and Jiaxiong Qiu and Zheng Zhu and Guan Huang and Zhizhong Su},
+      year={2025},
+      eprint={2505.23171},
+      archivePrefix={arXiv},
+      primaryClass={cs.CV},
+      url={https://arxiv.org/abs/2505.23171},
+}
+```