| Video | Role | Length | Tracked | Lines up |
|---|
More settings
Palm Orientation, Explained
| Video | Role | Length | Tracked | Lines up |
|---|
Each camera gets its own hand tracker and its own tile, side by side. Motion capture records them all together, on one clock.
Pick the window to track, for example OptiTrack Motive with a camera's view showing. Motive keeps the cameras to itself, so this is how Hand Tracker can use them while Motive runs.
Drag a box around the part to track, such as one camera's view, or track the whole window.
Watch a capture rig's camera live and track the hands in it, as your capture-fleet dashboard shows it. Enter the dashboard's address once; it's kept on this computer.
Pick a camera to watch it and track its hands; pick it again to go back to your camera. Rigs recording now are listed first, and can always be watched. The pictures are the dashboard's previews (full-size keyframes where the camera has them): a new one every second or two, so hands jump between pictures. Nothing on the rig is changed.
Track recorded sessions from your capture-operations dashboard. Enter its address once; Hand Tracker reads how to reach its data from the site itself and keeps that on this computer.
Sign in with your dashboard account.
Or here:
With an OAK camera (OAK-D, OAK-D Lite, OAK-1…) the hands are found on the camera itself, and an OAK-D also measures each hand's distance. It needs a one-time setup: Hand Tracker downloads its own copy of Python 3.12 with Luxonis's depthai library, OpenCV and NumPy, and the camera models (about 150 MB, into its own folder; nothing else on your computer changes). The tracking code comes from geaxgx/depthai_hand_tracker.