ONNX models required for face detection and recognition. These files are not included in the repository — download them manually.

• Source: insightface/detection/scrfd
• Download: https://github.com/deepinsight/insightface/releases
• File: det_10g.onnx (16.1 MB)
• Input: [1, 3, 640, 640] float32 (NCHW, normalized)
• Output: 9 tensors (3 strides × scores/bboxes/landmarks)

• Source: insightface/recognition/arcface
• Download: https://github.com/deepinsight/insightface/releases
• File: w600k_r50.onnx (166 MB)
• Input: [1, 3, 112, 112] float32 (NCHW, normalized)
• Output: [1, 512] float32 embedding

# Create model directory
mkdir -p ~/.local/share/visage/models/

# Place downloaded models
cp det_10g.onnx ~/.local/share/visage/models/
cp w600k_r50.onnx ~/.local/share/visage/models/

Visage will look for models in $XDG_DATA_HOME/visage/models/ (defaults to ~/.local/share/visage/models/). Model paths can also be specified explicitly via CLI arguments or daemon configuration.

Verify downloads:

sha256sum ~/.local/share/visage/models/det_10g.onnx
sha256sum ~/.local/share/visage/models/w600k_r50.onnx