180 episodes of robot evaluation data created using the LeRobot framework. The dataset contains 73700 frames of video data at 30 frames per second, organized into 3 tasks. It was uploaded by user csacan to Hugging Face in May 2026.
Use Cases
- Training robot control models based on the provided video episodes.
- Evaluating robot performance on the three defined tasks.
- Benchmarking reinforcement learning algorithms using the structured episode data.
- Analyzing robot behavior from the first-person perspective (POV) video frames.
Strengths
- Contains 73700 frames of video data, providing substantial visual information.
- Includes 180 episodes across 3 distinct tasks for structured evaluation.
- Data is chunked into manageable files of 1000 frames each.
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
- LeRobot framework.
- Collection Method
- Created for evaluation purposes.