RoboTrack24
RoboTrack Real v3
An evaluation dataset of 318 real-world video clips with sparse 2D point trajectories and visibility annotations for point tracking in robotic scenes. Each example includes the original RGB video, a rendered visualization of point tracks, and NumPy annotation archives.
Downloads15
Episodes318
Why This Matters for Physical AI
This dataset provides dense 2D point trajectory annotations on real robotic manipulation videos, enabling training and evaluation of point tracking models that are essential for visual perception and motion understanding in physical AI systems.
Technical Profile
- Modalities
- rgbpoint_track
- Task Types
- point_tracking
- Episodes
- 318
- Data Format
- video (H.264 MP4) + NumPy NPZ
- Annotation Types
- point_trajectoriesvisibility_labels
Access
Need custom rgb data?
Claru builds purpose-built datasets for any environment applications with dense human annotations and quality assurance.
Request a Sample Pack