Robomimic Lift: 500 Robot Manipulation Episodes for Imitation Learning

500 episodes of robot manipulation data were created using the LeRobot framework. The dataset contains 40,793 total frames recorded at 10 frames per second, focusing on a single task performed by a Panda robot. It was authored by yananchen and last updated on December 1, 2025.

Use Cases

• Training imitation learning agents based on recorded robot demonstrations.
• Developing and benchmarking robotic control policies based on the single-task structure.
• Analyzing robot motion and state transitions based on the 10 fps video and sensor data.
• Creating synthetic training data for reinforcement learning based on the provided episodes.

Strengths

• 500 total episodes provide a substantial collection of demonstrations.
• 40,793 frames captured at 10 fps offer detailed temporal resolution.
• Data is structured into chunks of 1000 episodes for manageable processing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset's license is unspecified, requiring verification before use.

Provenance

Source

LeRobot framework

Collection Method

Created using robot demonstrations.

Time Range

null

Freshness

Last updated 2025-12-01 21:49:05; freshness should be verified.

Geography

null