First-person capture that records what the demonstrator sees and how they move — RGB, depth, and motion — on the same clock as touch.
What it captures
First-person RGB aligned to the demonstrator's view, with stable exposure for long runs.
Stereo depth recovered from disparity for geometry, reach, and clutter around the hands.
IMU dynamics — acceleration, angular rates, and movement cues for inertia, rhythm, and effort.
Global-shutter stereo at 4000×1200 @ 30 fps — 1/2.6″ sensor, 3.0 µm pixels, wireless over WiFi with onboard compute and microSD.
One clock, every modality
The rig shares a single timebase with the glove and skin, so first-person video, depth, motion, and touch line up frame by frame — the alignment that makes episodes trainable.