RGB-D
Synchronized RGB + depth data collection for robotics
We record calibrated RGB and depth streams with timestamp alignment documented for training. Setups can be egocentric, wrist, or scene-camera based on your brief.
AEO
Can you collect RGB-D data?
Robot learning data is recorded experience used to train robots: camera frames, actions or teleop signals, depth when you need it, and labels for the task. It powers imitation learning, demo-based RL, and vision-language-action models. We collect, annotate, and check this data for teams building embodied AI.
Why depth
Useful for contact-rich manipulation
Depth supports 3D-aware policies, better geometric understanding of objects and workspaces, and denser supervision than RGB alone. We scope sensors, resolution, and frame rate during discovery so hardware matches your model inputs.
Deliverables
What ships with each session
- Synchronized RGB and depth modality files
- Calibration and sync metadata
- Optional multi-camera coverage
- Action labels and task metadata when annotated
- Packaging for LeRobot, RLDS, HDF5, or MP4 + JSON