G1 Brainco Robot Toast Manipulation Recordings

41865 video frames at 30 fps capture a Unitree G1 Brainco humanoid robot performing a single task. The dataset contains 101 episodes and was created by n1robotics using the LeRobot framework. It was last updated on the Hugging Face platform in April 2026.

Use Cases

• Train visual imitation learning models using the 30 fps video frames for robot manipulation.
• Develop policy learning algorithms from the 101 recorded episodes of a single, repeated task.
• Benchmark robotic control systems using the high-frequency (30 fps) video data of a Unitree G1 robot.
• Analyze robot motion and task execution consistency across the 101 provided episodes.

Strengths

• Contains 41865 individual video frames for detailed motion analysis.
• Offers 101 distinct episodes of task execution for training robust models.
• Data captured at a standard video frame rate of 30 frames per second.

Limitations

• Limited to a single task type, reducing generalizability to other robotic actions.
• No information on data diversity, sensor types beyond video, or annotation quality.
• Dataset size is moderate (100 MB for data, 500 MB for video) compared to large-scale robotics datasets.

Provenance

Source

n1robotics on Hugging Face.

Collection Method

Created using the LeRobot framework for robotics data collection.

Freshness

Last updated in April 2026.