Stereo & Multi-Camera Egocentric
Head-mounted stereo, wrist cameras, calibration and 3D head and hand trajectories.
Demo: stereo pair on top, left and right wrist cameras below.
Overview
Depth is what monocular egocentric video lacks. These sessions are recorded with a head-mounted stereo pair, often alongside left and right wrist cameras, so you get metric 3D hand positions and close-up views of every grasp.
The largest set (50,000 hours, about 70% homes and 30% commercial sites) ships with stereo calibration and head and hand trajectories in world and head-relative coordinates. A 5,000-hour set covers factory work such as wiring-harness and electrical component assembly (60% industrial, 40% homes), and a 100-hour set adds full left and right wrist streams.
Additional multi-camera sessions combine stereo, wrist cameras and fingertip force sensing; volume for these is confirmed per project.
Environments
- Homes
- Retail and commercial sites
- Electrical and wiring-harness assembly
- Electronics workbenches
Built for
- 3D hand pose and hand–object interaction
- Bimanual policy learning with wrist views
- Depth-aware world models
- Retargeting human hand trajectories to robot hands
Scenarios
PreviewFolding Clothes, Stereo + Both Wrists
Folding a shirt on a table, seen by a head-mounted stereo pair and a camera on each wrist.
PreviewElectronic Assembly, Stereo + Both Wrists
Fitting small electronic parts at a white workbench with gloved hands, recorded in stereo with both wrist cameras.
PreviewTabletop Sorting with 3D Hand Pose
Sorting items on a tray, shown as left and right stereo views with reprojected hand keypoints and the reconstructed 3D hands.
Capture specifications
Originals are 4K or 1080p at 30 fps per camera, with stereo calibration and head and hand trajectories (world and relative). Demo clips are downscaled composites: stereo pair on top, wrist cameras below.
| Scenario | Resolution | FPS | Sample length | Device | Extras |
|---|---|---|---|---|---|
| Folding Clothes, Stereo + Both Wrists | 4K / 1080p per camera | 60 | 20s | Stereo head rig + 2 wrist cameras | H.264 |
| Electronic Assembly, Stereo + Both Wrists | 4K / 1080p per camera | 60 | 20s | Stereo head rig + 2 wrist cameras | H.264 |
| Tabletop Sorting with 3D Hand Pose | 4K / 1080p per camera | 30 | 1m | Head-mounted stereo rig | H.264 · annotation demo |
Licensing & availability
| Inventory | 55,100 hours · indicative — exact hours confirmed in writing in your quote |
|---|---|
| Non-exclusive licence | Yes — default for all in-stock hours |
| Exclusive licence | On request, subject to existing licences |
| Commercial model training | Included |
| Custom top-up collection | Available when you need more than stock or a new scenario |
| Annotation | 3D head and hand trajectories and calibration included on the main set. Further labels quoted. |
| Consent | Written consent from every person filmed; documentation available during licence review |
| Privacy | Samples de-identified before publication; delivery screening per licence terms |
| Delivery | Per-camera MP4 or MCAP, calibration JSON and trajectory CSV, to S3 / GCS / Azure or physical drives |
FAQ
What comes with each session?
Left and right head-camera video, stereo calibration, and head and hand trajectories in world and head-relative coordinates. Wrist-camera streams are included where recorded.
How is it priced?
Per project, by hours, camera streams and exclusivity.
