fernandotonon/QtMeshEditor-poselandmarks-onnx

BlazePose full landmark model (ONNX)

Google MediaPipe's Pose Landmarks Detector (BlazePose full), converted to ONNX.

• Input [1,256,256,3] RGB in [0,1] — the rotated, cropped person ROI.
• Outputs [1,195] = 39 × (x,y,z,visibility,presence) screen landmarks in 256-crop pixels; [1,1] pose-presence probability; [1,256,256,1] segmentation; [1,64,64,39] heatmap; [1,117] = 39 × (x,y,z) WORLD landmarks in metres, hip-centred (the input to the analytic body-pose / IK solver — landmarks 0–32 are the 33 real body joints).

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.