rkothari3/il_gym_pickup

LeRobot Dataset: Pick Cube Task

This is a LeRobot dataset for the Panda robot pick-and-place task in MuJoCo simulation.

Dataset Information

• Task: Pick cube and place at different location
• Robot: Panda manipulator (gym-hil simulation)
• Total Episodes: 7
• Total Frames: 348
• FPS: 10
• Cameras: front view, wrist view
• Codebase Version: v3.0
• Action Space: 4D (x, y, z, gripper)
• Observation Space: RGB images (128x128) + robot state (18D)

Episode Lengths

• Episode 0: 38 frames (3.8s)
• Episode 1: 21 frames (2.1s)
• Episode 2: 31 frames (3.1s)
• Episode 3: 43 frames (4.3s)
• Episode 4: 50 frames (5.0s)
• Episode 5: 65 frames (6.5s)
• Episode 6: 100 frames (10.0s)

Loading the Dataset

This dataset uses the LeRobot format. To load it:

from lerobot.datasets.lerobot_dataset import LeRobotDataset

# Load dataset
dataset = LeRobotDataset("rkothari3/il_gym_pickup")

print(f"Episodes: {dataset.num_episodes}")
print(f"Frames: {len(dataset)}")

# Access a sample
sample = dataset[0]
print(f"Action: {sample['action']}")
print(f"Front camera: {sample['observation.images.front'].shape}")
print(f"Wrist camera: {sample['observation.images.wrist'].shape}")
print(f"State: {sample['observation.state'].shape}")

Training a Policy

Train a policy using this dataset:

# Diffusion Policy
lerobot-train \
  --policy.type=diffusion \
  --dataset.repo_id=rkothari3/il_gym_pickup \
  --env.type=gym_hil \
  --env.task=PandaPickCubeKeyboard-v0

# ACT Policy
lerobot-train \
  --policy.type=act \
  --dataset.repo_id=rkothari3/il_gym_pickup \
  --env.type=gym_hil \
  --env.task=PandaPickCubeKeyboard-v0

Dataset Structure

The dataset follows LeRobot's standard format:

• data/: Parquet files containing observations, actions, and metadata
• meta/: Dataset metadata including episode information and statistics
• videos/: MP4 video files for camera observations (AV1 codec)

Note: This dataset cannot be loaded with the standard HuggingFace datasets library. Use the LeRobot library instead.

Citation

If you use this dataset, please cite the LeRobot project:

@misc{cadene2024lerobot,
    author = {Cadene, Remi and Alibert, Simon and Soare, Alexander and Gallouedec, Quentin and Zouitine, Adil and others},
    title = {LeRobot: State-of-the-art Machine Learning for Real-World Robotics in Pytorch},
    howpublished = "\url{https://github.com/huggingface/lerobot}",
    year = {2024}
}