LumosRobotics commited on
Commit
156beac
·
verified ·
1 Parent(s): d891b27

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +99 -102
README.md CHANGED
@@ -5,13 +5,13 @@ language:
5
  tags:
6
  - robotics
7
  - manipulation
8
- - multimodal
9
  - trajectory-data
10
- - vision-sensors
 
 
11
  license: other
12
  task_categories:
13
  - robotics
14
- multimodal: vision+action
15
  dataset_info:
16
  features:
17
  - name: rgb_images
@@ -37,93 +37,62 @@ dataset_info:
37
  data_files: "**/*"
38
  ---
39
 
40
- <h1 align="center" style="font-size: 40px; font-weight: bold;">
41
- FastUMI Pro™ Robotics Dataset
 
 
42
  </h1>
43
 
44
- <h3 align="center">
45
- Enterprise-Grade Data Engine for Embodied AI
 
46
  </h3>
47
 
48
- <p align="center">
49
- <img src="https://img.shields.io/badge/Product-FastUMI_Pro-brightgreen?style=flat"/>
50
- <img src="https://img.shields.io/badge/Multimodal-7_Sensors-orange?style=flat"/>
51
- <img src="https://img.shields.io/badge/Trajectory_Data-150K-blue?style=flat"/>
52
- <img src="https://img.shields.io/badge/Application-VLA_Training-purple?style=flat"/>
53
- </p>
54
-
55
- ---
56
 
57
- ## 📖 Overview
 
 
58
 
59
- The **FastUMI Pro Sample Dataset** contains a small number of demonstration trajectories
60
- (only **dozens of episodes**, *not* a large-scale dataset).
61
- It showcases the multimodal sensing capabilities of the FastUMI Pro system, including:
62
 
63
- - RGB camera streams
64
- - Visual SLAM pose trajectories
65
- - Vive tracking data
66
- - Time-of-Flight (ToF) point clouds
67
- - Clamp (gripper gap) measurements
68
- - Fused pose trajectories
69
 
70
- This dataset is intended as a **public preview** of the data modality, structure, and quality of FastUMI Pro.
71
- For full-scale datasets or customized collection services, please **contact the FastUMI team directly**.
72
 
73
  ---
74
 
75
- ## Key Features
76
 
77
- - **High-precision spatial tracking**
78
- - **Multi-sensor synchronization** across RGB, SLAM, Vive, ToF, and clamp channels
79
- - **Standardized directory and timestamp structure**
80
- - **Ready for embodied AI, imitation learning, and robotics research**
81
- - **Hardware-agnostic data format** for cross-platform manipulation applications
82
 
83
- ---
84
 
85
- <!-- **VLA Model Results**: [TBD]
 
 
 
86
 
87
- ## 🛠️ Toolchain
88
- | Tool | Description | Link |
89
- |------|-------------|------|
90
- | **Single-Arm Demo Replay** | Single-arm data replay code | [GitHub](https://github.com/Loki-Lu/FastUMI_replay_singleARM) |
91
- | **Dual-Arm Demo Replay** | Dual-arm data replay code | [GitHub](https://github.com/Loki-Lu/FastUMI_replay_dualARM) |
92
- | **Hardware SDK** | FastUMI hardware development kit | [GitHub](https://github.com/FastUMIRobotics/FastUMI_Hardware_SDK) |
93
- | **Monitor Tool** | Real-time device monitoring | [GitHub](https://github.com/FastUMIRobotics/FastUMI_Monitor_Tool) |
94
- | **Data Collection** | Data collection utilities | [GitHub](https://github.com/FastUMIRobotics/FastUMI_Data_Collection) |
95
- -->
96
 
97
- ---
 
 
 
 
 
98
 
99
- <!-- ### Research & Applications
100
- - **Paper**: [MLM: Learning Multi-task Loco-Manipulation Whole-Body Control for Quadruped Robot with Arm](https://arxiv.org/abs/2508.10538)
101
- - **Tutorial**: PI0 (FastUMI Data Lightweight Adaptation, Version V0) Full Pipeline
102
- -->
103
 
104
  ---
105
 
106
- ## 📥 Data Download
107
-
108
- ### Example Dataset
109
- ```bash
110
- # Direct download (may be slow in some regions)
111
- huggingface-cli download FastUMIPro/example_data_fastumi_pro_raw --repo-type dataset --local-dir ~/fastumi_data/
112
- ```
113
 
114
- Mirror Download (Recommended)
115
- ```bash
116
- # Set mirror endpoint
117
- export HF_ENDPOINT=https://hf-mirror.com
118
- ```
119
 
120
- # Download via mirror
121
- huggingface-cli download --repo-type dataset --resume-download FastUMIPro/example_data_fastumi_pro_raw --local-dir ~/fastumi_data/
122
- 📁 Data Structure
123
- Each session represents an independent operation "episode" containing observation data and action sequences.
124
- ```
125
- Directory Structure
126
- text
127
  session_001/
128
  └── device_label_xv_serial/
129
  └── session_timestamp/
@@ -147,47 +116,75 @@ session_001/
147
  └── merge_stats.txt
148
  ```
149
 
150
- ## Data Specifications
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
151
 
152
- | Data Type | Path | Shape| Type | Description |
153
- | :--- | :--- | :--- | :--- | :--- |
154
- | **RGB Images** | `session_XXX/RGB_Images/Video.MP4` | `(frames, 1080, 1920, 3)`| `uint8`| Camera video data, 60 FPS |
155
- | **SLAM Poses** | `session_XXX/SLAM_Poses/slam_raw.txt` | `(timestamps, 7)`| `float` | UMI end-effector poses |
156
- | **Vive Poses** | `session_XXX/Vive_Poses/vive_data_tum.txt` | `(timestamps, 7)`| `float` | Vive base station poses |
157
- | **ToF PointClouds** | `session_XXX/PointClouds/pointcloud_...pcd` | `pcd format` | pcd | Time-of-Flight point cloud data |
158
- | **Clamp Data** | `session_XXX/Clamp_Data/clamp_data_tum.txt` | `(timestamps, 1)`| `float` | Gripper spacing (mm) |
159
- | **Merged Trajectory** | `session_XXX/Merged_Trajectory/merged_trajectory.txt` | `(timestamps, 7)`| `float` | Fused trajectory (Vive/UMI based on velocity) |
160
 
161
- ### Pose Data Format
 
162
 
163
- All pose data (SLAM, Vive, Merged) follow the same format:
164
 
165
- | Data | Description |
166
- | :--- | :--- |
167
- | **Timestamp** | Unix timestamp of the trajectory data |
168
- | **Pos X** | X-coordinate of position (meters) |
169
- | **Pos Y** | Y-coordinate of position (meters) |
170
- | **Pos Z** | Z-coordinate of position (meters) |
171
- | **Q_X** | X-component of orientation quaternion |
172
- | **Q_Y** | Y-component of orientation quaternion |
173
- | **Q_Z** | Z-component of orientation quaternion |
174
- | **Q_W** | W-component of orientation quaternion |
175
 
176
- ## 🔄 Data Conversion
177
- [TBD]
178
 
179
- ## 🤝 Collaboration
180
- FastUMI Pro dataset is available for research collaboration. The full FastUMI-150K dataset has been provided to partner research teams for large-scale model training.
181
 
182
  ## 📞 Contact
183
- ***
184
 
185
- > ### ☎️ 开发团队联系方式
186
- >
187
- > 对于任何问题或建议,请随时联系我们的开发团队。
188
- > | **负责人 (Lead)** | Ding Yan |
189
- > | :--- | :--- |
190
- > | **Email** | **[dingyan@lumosbot.tech](mailto:dingyan@lumosbot.tech)** |
191
- > | **WeChat** | **`Duke_dingyan`** |
192
 
193
- ***
 
5
  tags:
6
  - robotics
7
  - manipulation
 
8
  - trajectory-data
9
+ - multimodal
10
+ - embodied-ai
11
+ multimodal: vision+language+action
12
  license: other
13
  task_categories:
14
  - robotics
 
15
  dataset_info:
16
  features:
17
  - name: rgb_images
 
37
  data_files: "**/*"
38
  ---
39
 
40
+ <div align="center">
41
+
42
+ <h1 style="font-size: 42px; font-weight: 900; margin-bottom: 10px;">
43
+ FastUMI Pro – Multimodal Sample Dataset
44
  </h1>
45
 
46
+ <h3 style="font-size: 20px; font-weight: 400; margin-top: -10px;">
47
+ Small-Scale Demonstration Data From the FastUMI Pro Multimodal Sensing System
48
+ <br>(Only Dozens of Trajectories — Full Dataset Available Upon Request)
49
  </h3>
50
 
51
+ <br>
 
 
 
 
 
 
 
52
 
53
+ <img src="https://img.shields.io/badge/FastUMI-Pro-brightgreen" />
54
+ <img src="https://img.shields.io/badge/Sample%20Dataset-Small-blue" />
55
+ <img src="https://img.shields.io/badge/Multimodal-Vision%20%7C%20Pose%20%7C%20PointCloud-orange" />
56
 
57
+ <br><br>
 
 
58
 
59
+ <a href="https://fastumi.com/pro/">Project Homepage</a>
 
 
 
 
 
60
 
61
+ </div>
 
62
 
63
  ---
64
 
65
+ ## 📖 Overview
66
 
67
+ The **FastUMI Pro Sample Dataset** provides a publicly available preview of the multimodal sensing pipeline used by the FastUMI Pro system.
 
 
 
 
68
 
69
+ This release contains **only a small number of demonstration trajectories**, intended for:
70
 
71
+ - System testing
72
+ - Algorithm prototyping
73
+ - Pipeline integration
74
+ - Demonstration of multimodal alignment quality
75
 
76
+ This dataset includes:
 
 
 
 
 
 
 
 
77
 
78
+ - RGB camera frames
79
+ - Visual SLAM trajectories
80
+ - Vive tracking poses
81
+ - ToF point clouds
82
+ - Clamp (gripper width) sensor readings
83
+ - Fused multi-sensor trajectory data
84
 
85
+ Full-scale datasets and enterprise data services are available upon request.
 
 
 
86
 
87
  ---
88
 
89
+ ## 📁 Directory Structure
 
 
 
 
 
 
90
 
91
+ <div align="center">
92
+ <h3>Dataset Folder Layout</h3>
93
+ </div>
 
 
94
 
95
+ ```markdown
 
 
 
 
 
 
96
  session_001/
97
  └── device_label_xv_serial/
98
  └── session_timestamp/
 
116
  └── merge_stats.txt
117
  ```
118
 
119
+ ---
120
+
121
+ ## 📊 Data Specifications
122
+
123
+ | **Data Type** | **Path** | **Shape** | **Type** | **Description** |
124
+ |--------------|----------|-----------|----------|-----------------|
125
+ | RGB Images | RGB_Images/Frames/*.jpg | (H, W, 3) | uint8 | Multi-view RGB images |
126
+ | SLAM Poses | SLAM_Poses/slam_raw.txt | (N, 8) | float | Timestamp + SE(3) pose |
127
+ | Vive Poses | Vive_Poses/vive_data_tum.txt | (N, 8) | float | Vive tracking |
128
+ | ToF PointClouds | ToF_PointClouds/PointClouds/*.pcd | variable | pcd | ToF 3D point clouds |
129
+ | Clamp Data | Clamp_Data/clamp_data_tum.txt | (N, 2) | float | Timestamp + clamp width |
130
+ | Merged Trajectory | Merged_Trajectory/merged_trajectory.txt | (N, 8) | float | Fused pose trajectory |
131
+
132
+ ---
133
+
134
+ ## 🧭 Pose Data Format
135
+
136
+ ```markdown
137
+ timestamp x y z qx qy qz qw
138
+ ```
139
+
140
+ | Field | Description |
141
+ |-------|-------------|
142
+ | timestamp | Unix timestamp |
143
+ | x | Position X (meters) |
144
+ | y | Position Y (meters) |
145
+ | z | Position Z (meters) |
146
+ | qx | Quaternion X |
147
+ | qy | Quaternion Y |
148
+ | qz | Quaternion Z |
149
+ | qw | Quaternion W |
150
+
151
+ ---
152
+
153
+ ## 📥 Download
154
+
155
+ ```bash
156
+ huggingface-cli download FastUMIPro/example_data_fastumi_pro_raw \
157
+ --repo-type dataset \
158
+ --local-dir ./fastumi_sample/
159
+ ```
160
+
161
+ Mirror (optional):
162
+
163
+ ```bash
164
+ export HF_ENDPOINT=https://hf-mirror.com
165
+ ```
166
+
167
+ ---
168
 
169
+ ## ⚠️ Dataset Scale
 
 
 
 
 
 
 
170
 
171
+ This dataset contains **only a small number of sample episodes**
172
+ and is **not intended for large-scale model training**.
173
 
174
+ For:
175
 
176
+ - Full datasets
177
+ - Enterprise multimodal collection
178
+ - Research collaboration
 
 
 
 
 
 
 
179
 
180
+ Please contact the FastUMI team.
 
181
 
182
+ ---
 
183
 
184
  ## 📞 Contact
 
185
 
186
+ Lead: **Ding Yan**
187
+ Email: **dingyan@lumosbot.tech**
188
+ WeChat: **Duke_dingyan**
 
 
 
 
189
 
190
+ ---