A dataset containing 100 episodes of robot action data, totaling 103,706 frames and 300 videos, created using LeRobot. It was uploaded by YOLO2431 to Hugging Face on May 7, 2026. The dataset is structured for a single task involving a Yam bimanual robot.
Use Cases
- Training robot action prediction models based on recorded episodes
- Developing control policies for bimanual robots based on task-specific data
- Benchmarking reinforcement learning algorithms on a defined manipulation task
- Analyzing robot performance across 100 episodes of a single task
Strengths
- Contains 100 episodes of task-specific robot data
- Includes 103,706 frames and 300 videos for visual analysis
- Dataset structure is defined with a single task and splits for training
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
- Freshness
- Last updated 2026-05-07 03:33:40; freshness should be verified