Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,28 +1,24 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
task_categories:
|
| 4 |
-
|
| 5 |
tags:
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
configs:
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
---
|
| 18 |
|
| 19 |
# SO-101 Pick Cube Dataset
|
| 20 |
|
| 21 |
-
|
| 22 |
-
<img src="preview.png" alt="Dataset Preview" width="800"/>
|
| 23 |
-
</p>
|
| 24 |
-
|
| 25 |
-
A robotics manipulation dataset featuring the **SO-101 robot arm** performing pick-and-place tasks in MuJoCo simulation. The robot learns to pick up a cube and place it into a bin.
|
| 26 |
|
| 27 |
## Dataset Details
|
| 28 |
|
|
@@ -30,68 +26,42 @@ A robotics manipulation dataset featuring the **SO-101 robot arm** performing pi
|
|
| 30 |
|----------|-------|
|
| 31 |
| Robot | SO-101 (6-DoF arm with gripper) |
|
| 32 |
| Task | Pick cube and place in bin |
|
| 33 |
-
| Episodes | 993 |
|
|
|
|
| 34 |
| FPS | 30 |
|
| 35 |
| Cameras | 3 (front, overhead, wrist) |
|
| 36 |
| Resolution | 640x480 |
|
| 37 |
-
| Format | LeRobot
|
| 38 |
|
| 39 |
## Cameras
|
| 40 |
|
| 41 |
-
- **Front Camera**:
|
| 42 |
-
- **Overhead Camera**: Top-down
|
| 43 |
-
- **Wrist Camera**:
|
| 44 |
|
| 45 |
-
##
|
| 46 |
|
| 47 |
-
|
| 48 |
-
- `shoulder_pan` - Base rotation
|
| 49 |
-
- `shoulder_lift` - Shoulder elevation
|
| 50 |
-
- `elbow_flex` - Elbow bend
|
| 51 |
-
- `wrist_flex` - Wrist pitch
|
| 52 |
-
- `wrist_roll` - Wrist rotation
|
| 53 |
-
- `gripper` - Gripper open/close
|
| 54 |
|
| 55 |
-
## Usage
|
| 56 |
|
| 57 |
```python
|
| 58 |
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
|
| 59 |
|
| 60 |
dataset = LeRobotDataset("gpudad/so101_pick_cube")
|
| 61 |
-
print(f"
|
| 62 |
-
print(f"
|
| 63 |
```
|
| 64 |
|
| 65 |
-
##
|
| 66 |
|
| 67 |
```bash
|
| 68 |
-
# Train ACT policy
|
| 69 |
python lerobot/scripts/train.py \
|
| 70 |
--dataset.repo_id=gpudad/so101_pick_cube \
|
| 71 |
-
--policy=act \
|
| 72 |
-
--
|
| 73 |
```
|
| 74 |
|
| 75 |
-
## Environment
|
| 76 |
-
|
| 77 |
-
The simulation environment uses MuJoCo with:
|
| 78 |
-
- SO-101 robot arm mounted on a table
|
| 79 |
-
- A colored cube as the manipulation target
|
| 80 |
-
- A bin as the placement destination
|
| 81 |
-
- Randomized cube positions for diverse training data
|
| 82 |
-
|
| 83 |
## License
|
| 84 |
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
## Citation
|
| 88 |
-
|
| 89 |
-
```bibtex
|
| 90 |
-
@dataset{so101_pick_cube_2025,
|
| 91 |
-
title={SO-101 Pick Cube Dataset},
|
| 92 |
-
author={gpudad},
|
| 93 |
-
year={2025},
|
| 94 |
-
publisher={Hugging Face},
|
| 95 |
-
url={https://huggingface.co/datasets/gpudad/so101_pick_cube}
|
| 96 |
-
}
|
| 97 |
-
```
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
task_categories:
|
| 4 |
+
- robotics
|
| 5 |
tags:
|
| 6 |
+
- lerobot
|
| 7 |
+
- so101
|
| 8 |
+
- manipulation
|
| 9 |
+
- pick-and-place
|
| 10 |
+
- simulation
|
| 11 |
+
- mujoco
|
| 12 |
configs:
|
| 13 |
+
- config_name: default
|
| 14 |
+
data_files:
|
| 15 |
+
- split: train
|
| 16 |
+
path: data/**/*.parquet
|
| 17 |
---
|
| 18 |
|
| 19 |
# SO-101 Pick Cube Dataset
|
| 20 |
|
| 21 |
+
A large-scale robotics manipulation dataset featuring the SO-101 robot arm performing pick-and-place tasks in MuJoCo simulation. The robot learns to pick up a cube and place it into a bin.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
|
| 23 |
## Dataset Details
|
| 24 |
|
|
|
|
| 26 |
|----------|-------|
|
| 27 |
| Robot | SO-101 (6-DoF arm with gripper) |
|
| 28 |
| Task | Pick cube and place in bin |
|
| 29 |
+
| Episodes | 10,993 |
|
| 30 |
+
| Total Frames | 1,456,901 |
|
| 31 |
| FPS | 30 |
|
| 32 |
| Cameras | 3 (front, overhead, wrist) |
|
| 33 |
| Resolution | 640x480 |
|
| 34 |
+
| Format | LeRobot v3.0 |
|
| 35 |
|
| 36 |
## Cameras
|
| 37 |
|
| 38 |
+
- **Front Camera**: External view of the robot and workspace
|
| 39 |
+
- **Overhead Camera**: Top-down view of the workspace
|
| 40 |
+
- **Wrist Camera**: Mounted on the robot's wrist for close-up manipulation view
|
| 41 |
|
| 42 |
+
## State/Action Space
|
| 43 |
|
| 44 |
+
6 joints: `shoulder_pan`, `shoulder_lift`, `elbow_flex`, `wrist_flex`, `wrist_roll`, `gripper`
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 45 |
|
| 46 |
+
## Usage with LeRobot
|
| 47 |
|
| 48 |
```python
|
| 49 |
from lerobot.common.datasets.lerobot_dataset import LeRobotDataset
|
| 50 |
|
| 51 |
dataset = LeRobotDataset("gpudad/so101_pick_cube")
|
| 52 |
+
print(f"Episodes: {dataset.num_episodes}")
|
| 53 |
+
print(f"Frames: {dataset.num_frames}")
|
| 54 |
```
|
| 55 |
|
| 56 |
+
## Training
|
| 57 |
|
| 58 |
```bash
|
|
|
|
| 59 |
python lerobot/scripts/train.py \
|
| 60 |
--dataset.repo_id=gpudad/so101_pick_cube \
|
| 61 |
+
--policy.type=act \
|
| 62 |
+
--env.type=so101_pick_cube
|
| 63 |
```
|
| 64 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
## License
|
| 66 |
|
| 67 |
+
Apache 2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|