Unitree Box Move Blue Full: Robot Teleoperation Data for the Unitree G1

550 episodes of robot teleoperation data for the Unitree G1 robot, totaling 475,206 frames. The dataset was created using the LeRobot framework and published by author nepyope on Hugging Face, with a last recorded update in February 2026. It contains a single task focused on moving a blue box.

Use Cases

• Training imitation learning models based on the 550 recorded teleoperation episodes.
• Developing robot control policies for box manipulation tasks using the provided state and action data.
• Benchmarking reinforcement learning algorithms on a real-world robot platform using the structured episodes.
• Analyzing robot motion and task performance from the 30 FPS video recordings likely associated with the data.

Strengths

• Contains 550 distinct task episodes, providing a substantial number of trials.
• Includes 475,206 total frames, suggesting detailed temporal resolution for analysis.
• Data is structured into chunks of 1000 frames, which may facilitate efficient data loading and processing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Created using the LeRobot framework, likely involving human teleoperation of a Unitree G1 robot.

Freshness

Last updated 2026-02-12 17:26:30; freshness should be verified.