Hand-Tracking Skeleton IK (2D)

Camera-based hand tracking is one of the core sensing problems in spatial computing: a depth or RGB camera only ever detects a handful of noisy fingertip points, yet the device needs a full, temporally stable hand skeleton to drive gestures. This 2D simulator runs the real pipeline live — inject adjustable sensor jitter, smooth it with an exponential moving average, and watch a FABRIK inverse-kinematics solver reconstruct a rigid 19-joint, 5-finger skeleton from just those five filtered targets every frame. A gesture classifier reads the solved geometry independently of the gesture you picked, showing exactly how a headset turns raw, noisy sensor data into a reliable pinch, point or fist. Drag to pan the view and scroll to zoom in on the joints.