Voxel512012cc-by-4.0
TUM RGB-D SLAM benchmark (FiftyOne multimodal MCAP)
47 sequences totalling 48m 12s of RGB-D video recorded with Microsoft Kinect sensors and motion-capture ground truth, organized into six benchmark categories for SLAM and visual odometry evaluation.
Downloads163
Episodes47
Hours0.8
Why This Matters for Physical AI
This dataset provides calibrated RGB-D sequences with precise motion-capture ground truth, establishing a standard benchmark for evaluating spatial perception and localization algorithms critical to embodied AI systems.
Technical Profile
- Modalities
- rgbdepthproprioceptionimu
- Robot Embodiments
- mobile_manipulator
- Environment
- labhomeoffice
- Task Types
- visual_odometryslam3d_reconstructionnavigation
- Episodes
- 47
- Total Hours
- 0.8
- Data Format
- MCAP
- Annotation Types
- ground_truth_posescamera_calibration
- License
- cc-by-4.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for lab applications with dense human annotations and quality assurance.
Request a Sample Pack