Webcam-driven hand-gesture file navigator using MediaPipe 21-point landmarks and PyAutoGUI OS actions
Hook up a webcam and navigate your file system entirely with hand gestures. repoNavTouch runs a See → Think → Act pipeline: OpenCV captures frames, MediaPipe detects 21 3-D hand landmarks per frame, a GestureRecogniser state machine classifies them into a Gesture enum (pinch, swipe-left/right, open-palm-up/down, two-fingers-up), and a FileNavigator maps each gesture to OS-level directory navigation actions via PyAutoGUI.
Gesture-based file navigation using real-time hand tracking.
• See: OpenCV webcam loop at configurable device index; MediaPipe Hands model detects 21 3-D landmarks per frame.
• Think: gestures.py geometry helpers + GestureRecogniser state machine; normalised landmark positions; configurable pinch threshold.
• Act: file_navigator.py maps gestures to OS keyboard/mouse events via PyAutoGUI; supports both file explorer and terminal navigation.
• Gestures: pinch = enter/open, swipe-left = back, swipe-right = forward, open-palm-up/down = scroll, two-fingers-up = cd ..
• HUD overlay: OpenCV window with live landmark visualisation and current gesture label.
• Tests: pytest unit tests for gesture classification and file navigation logic.