This is the same gesture recognition tech from my MEng project. Real-time hand tracking and gesture classification, running entirely in your browser via MediaPipe. No data leaves your device.
Show your hand to the camera. The model will detect and classify your gesture in real-time using 21 landmark points.
Video frames are captured from your webcam at 30-60fps and processed entirely on-device using WebGL acceleration.
MediaPipe Hands detects 21 3D hand keypoints — fingertips, joints, wrist — with sub-millimetre relative accuracy.
Joint angles and relative positions are computed. Gesture rules classify the hand state — the same approach used in my MEng smart glasses project.
Results render at 60fps with skeleton overlay. In production systems, these outputs trigger robot control, HMI events, or automation commands.