How zannbjir/kicau-mania works
This small Python application captures webcam frames, detects hand landmarks with MediaPipe, classifies a “Mikir” pose, and smooths detection across three frames. When the pose is stable, it plays the bundled cat video in a window; otherwise it closes that window and resets video playback. The README documents this user workflow. The sampled source is all in main.py, so the stages below are logical responsibilities within that file, not separate modules.
Main entry points: main.py Main loop · hand_landmarker.task Landmarker model · kicau.mp4 Cat video
Components 10
Other components
- User
- Main loop
main.py - Webcam
- Landmarker model
hand_landmarker.task - Hand detector
main.py - Pose classifier
main.py - Gesture smoother
main.py - Cat video
kicau.mp4 - Video playback
main.py - OpenCV windows
Connections 14
- User → Main loop: starts
- User → Main loop: presses key
- Main loop → Webcam: reads frames
- Main loop → Landmarker model: loads
- Main loop → Hand detector: configures
- Main loop → Hand detector: detects
- Hand detector → Pose classifier: landmarks
- Main loop → Pose classifier: classifies pose
- Pose classifier → Gesture smoother: gesture
- Main loop → Gesture smoother: updates history
- Gesture smoother → Video playback: dispatches
- Main loop → Cat video: reads
- Main loop → Video playback: reads frames
- Main loop → OpenCV windows: shows or closes