Egoinfinity: Derivative Scene Assets from Action100M Clips

Derivative scene assets for a curated subset of clips from the Action100M dataset. The preview version is built by the Rice RobotPI Lab and is not for general release. Its schema and contents may change.

Use Cases

• Train models for robot action retargeting based on the provided results for four robot embodiments.
• Analyze ego-centric scene assets derived from the Action100M dataset.
• Benchmark algorithms for action recognition using curated video clips.

Strengths

• Provides retargeting results for 104 specific clips on four robot embodiments.
• Built upon the large-scale Action100M dataset from Meta FAIR.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Rice-RobotPI-Lab

Collection Method

Derived from a curated subset of the Action100M (Meta FAIR) dataset.

Freshness

Last updated 2026-06-04 15:10:00; freshness should be verified.