fernandotonon/QtMeshEditor-blazeface-onnx

BlazeFace short-range face detector (ONNX)

Google MediaPipe's BlazeFace short-range detector, from the Face Landmarker .task bundle, converted to ONNX.

• Input input [1,128,128,3] RGB, normalized to [-1,1], keep-aspect letterboxed (zero border).
• Outputs regressors [1,896,16] (box + 6 keypoints per anchor), classificators [1,896,1] (score logits). SSD anchors: strides [8,16,16,16], min/max scale 0.1484375/0.75 → 896 anchors; decode + weighted NMS at IoU 0.3.
• Feeds the face ROI (rotation from the eye keypoints, 1.5× square-long) to the Face Mesh landmark model.

License

Apache-2.0 — this graph is a direct ONNX conversion of a Google MediaPipe model (Apache-2.0 code AND weights). Conversion + numerical-parity proof (vs the Python mediapipe reference): scripts/export-facecap-onnx.py, contract in docs/MOCAP_SPIKE.md.

How it is used

Mirror of one graph from fernandotonon/QtMeshEditor-models (mocap/…), which QtMeshEditor downloads on first use for its Performance Capture feature (video/webcam → facial morph + head + full-body skeletal animation, epic #869). This standalone repo is for discoverability; the app fetches from the aggregate repo.