gpudad/so101_pick_cube_v2

SO-101 Pick Cube Dataset

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.

Dataset Details

Property	Value
Robot	SO-101 (6-DoF arm with gripper)
Task	Pick cube and place in bin
Episodes	993
FPS	30
Cameras	3 (front, overhead, wrist)
Resolution	640x480
Format	LeRobot v2.0

Cameras

• Front Camera: Third-person view of the workspace
• Overhead Camera: Top-down bird's eye view
• Wrist Camera: First-person view from the gripper

Joint States

The dataset includes 6 joint positions and actions:

• shoulder_pan - Base rotation
• shoulder_lift - Shoulder elevation
• elbow_flex - Elbow bend
• wrist_flex - Wrist pitch
• wrist_roll - Wrist rotation
• gripper - Gripper open/close

Usage

from lerobot.common.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset("gpudad/so101_pick_cube")
print(f"Number of episodes: {dataset.num_episodes}")
print(f"Number of frames: {dataset.num_frames}")

Training with LeRobot

# Train ACT policy
python lerobot/scripts/train.py \
  --dataset.repo_id=gpudad/so101_pick_cube \
  --policy=act \
  --output_dir=outputs/act_so101_pick_cube

Environment

The simulation environment uses MuJoCo with:

• SO-101 robot arm mounted on a table
• A colored cube as the manipulation target
• A bin as the placement destination
• Randomized cube positions for diverse training data

License

This dataset is released under the Apache 2.0 License.

Citation

@dataset{so101_pick_cube_2025,
  title={SO-101 Pick Cube Dataset},
  author={gpudad},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/gpudad/so101_pick_cube}
}