Robot Manipulation Demonstrations with Synchronized RGB-D and Segmentation

LIBERO-Mem provides time-aligned, object-centric robot manipulation data. The dataset includes task-level metadata, synchronized RGB-D observations, segmentation maps, proprioception, and control actions stored in metainfo.json and .hdf5 files. It was created by libero-mem and last updated in November 2025.

Use Cases

• Train robot manipulation policies using synchronized RGB-D observations and control actions.
• Develop object-centric perception models with pixel-level segmentation maps.
• Analyze robot proprioception data in relation to bounding boxes and initial states from metadata.
• Benchmark robotic task performance using time-aligned demonstration files.

Strengths

• Includes synchronized multi-modal data streams: RGB-D, segmentation, proprioception, and actions.
• Provides object-centric and pixel-level representations for detailed analysis.
• Data is time-aligned across demonstration files and metadata.

Limitations

• Dataset size, row count, and specific file formats are not provided.
• Sample data is unavailable for previewing content structure.
• License and organization information are unknown, affecting usage clarity.

Provenance

Source

huggingface

Collection Method

Robot demonstrations, likely from simulation or physical robot trials.

Time Range

null

Freshness

Last updated November 2025.

Geography

null