Voxel51apache-2.0
RoboLab-EgoX (FiftyOne multimodal MCAP)
Complete RoboLab-EgoX corpus of 4,000 policy rollouts on NVIDIA's RoboLab manipulation benchmark, converted to native multimodal MCAP episodes with synchronized egocentric and exocentric camera views, depth streams, and whole-arm telemetry.
Downloads635
Episodes4000
Hours6.0
Likes1
Why This Matters for Physical AI
This dataset provides a large-scale benchmark of manipulation tasks with synchronized multimodal sensor streams and explicit success/failure labels, enabling training and evaluation of task-generalist robotic policies.
Technical Profile
- Modalities
- rgbdepthproprioceptionlanguage
- Action Space
- joint_positions
- Environment
- simulation
- Task Types
- manipulation
- Episodes
- 4000
- Total Hours
- 6.0
- Data Format
- MCAP
- Annotation Types
- language_instructionsreward_labels
- License
- apache-2.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack