Table Spill Cleanup Bimanual: A Multi-View Robot Manipulation Dataset

A human-in-the-loop, multi-view bimanual robot manipulation dataset for tabletop spill cleanup tasks. The dataset includes synchronized video, state and action trajectories, phase annotations, and success or failure labels, structured for compatibility with the LeRobot framework. It was created by ExylosAi and last updated on May 5, 2026.

Use Cases

• Training imitation learning models based on synchronized video and action trajectories.
• Developing and benchmarking robot policies for bimanual tabletop cleanup tasks.
• Analyzing task phases and failure modes using the provided phase and success labels.
• Researching human-in-the-loop robot learning from the described demonstration data.

Strengths

• Includes multiple synchronized data modalities: video, state/action trajectories, and annotations.
• Features specific task annotations for phases and success/failure outcomes.
• Structured for compatibility with the LeRobot robotics framework.

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

ExylosAi via Hugging Face Datasets.

Collection Method

Likely collected via human-in-the-loop robot demonstrations.

Freshness

Last updated 2026-05-05 15:55:16; freshness should be verified.