7,356 episodes of human teleoperation data for 65 atomic robot skills, comprising 1,495,313 video frames. This dataset is part of the larger 482-hour RoboCasa365 pretraining collection, mirrored in the LeRobot v3.0 layout for direct loading. The data was contributed by ember-lab-berkeley and was last updated in May 2026.
Use Cases
- Training robot imitation learning models based on human teleoperation demonstrations.
- Pretraining robotic foundation models on the described atomic single-skill tasks.
- Benchmarking robot skill generalization based on the 617 natural-language task phrasings.
- Analyzing multi-camera robot video data from the three 256x256 camera streams.
Strengths
- Contains 7,356 demonstration episodes, providing a substantial volume of training data.
- Covers 65 distinct atomic robot skills, offering a variety of manipulation tasks.
- Includes 1,495,313 video frames, which corresponds to approximately 21 hours of footage at 20 fps.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Freshness should be verified as the last update was in 2026.
Provenance
- Source
- ember-lab-berkeley
- Collection Method
- Human teleoperation recordings.
- Freshness
- Last updated 2026-05-11 02:29:44