Model details
Face Detection
Finds every face in a photo or live webcam feed and marks five landmarks per face.
- Model
- YuNet (OpenCV Zoo, 2023)
- Weights
- 53K parameters
- Input
- 1 × 3 × 640 × 640, BGR
- Output
- Boxes, scores, 5 landmarks
- Post-processing
- Anchor decoding + NMS in TypeScript
- Runtime
- ONNX Runtime Web, WASM