Chenwei1999cc-by-nc-sa-4.0
Benchmark-300 — Timestamped Cross-Domain Atomic Skill Calls
300 videos with fully timestamped skill calls and natural language descriptions across robot manipulation, cooking, and instructional domains, using a unified 41-skill library. Provides temporal grounding for atomic skills with adversarially verified annotations.
Downloads0
Episodes300
Why This Matters for Physical AI
This dataset enables temporal grounding of atomic skills from vision and language, supporting training of vision-language models for precise action understanding and temporal localization in embodied AI systems.
Technical Profile
- Modalities
- rgblanguage
- Robot Embodiments
- mobile_manipulator
- Action Space
- language
- Environment
- kitchenlabhome
- Task Types
- manipulationcookinggraspingpick_and_placepouringheating
- Episodes
- 300
- Data Format
- JSON
- Annotation Types
- language_instructionsaction_labelstemporal_groundingskill_calls
- License
- cc-by-nc-sa-4.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for kitchen applications with dense human annotations and quality assurance.
Request a Sample Pack