Voxel512024bsd-3-clause
NTNU Underwater Multi-Camera Visual-Inertial (FiftyOne multimodal MCAP)
A FiftyOne dataset with 8 underwater robot runs totalling 63m 19s of recording, featuring five monochrome cameras, two inertial units, and auxiliary sensors from the BlueROV2-based Ariel platform operating in the Trondheim Fjord and a pool laboratory.
Downloads151
Episodes8
Hours1.056
Why This Matters for Physical AI
This dataset provides multimodal underwater robot perception data with reference ground truth trajectories, enabling development and evaluation of visual-inertial odometry and SLAM algorithms for autonomous underwater vehicles in real-world marine environments.
Technical Profile
- Modalities
- rgbimubarometerrangefinderbattery_voltagethruster_outputs
- Robot Embodiments
- BlueROV2 Heavy (Ariel)
- Action Space
- thruster_outputs
- Environment
- outdoorlab
- Task Types
- visual-inertial-odometryslamnavigation
- Episodes
- 8
- Total Hours
- 1.056
- Data Format
- MCAP
- Annotation Types
- camera_calibrationreference_trajectoryframe_transforms
- License
- bsd-3-clause
Access
Need custom rgb data?
Claru builds purpose-built datasets for outdoor applications with dense human annotations and quality assurance.
Request a Sample Pack