MiniFASNet

MiniFASNet MXNet Gluon=>ONNX=>TensorRT

1.Reference

• MiniFASNet github: https://github.com/minivision-ai/Silent-Face-Anti-Spoofing

2.Export ONNX Model

• export 2.7_80x80_MiniFASNetV2 model

python3 export_onnx.py

• export 4_0_0_80x80_MiniFASNetV1SE model

python3 export_onnx.py --weights_file ./resources/anti_spoof_models/4_0_0_80x80_MiniFASNetV1SE.pth

3.Build MiniFASNet_trt Project

cd ../  # in project directory
mkdir build && cd build
cmake ..
make -j

4.Run MiniFASNet_trt

cd ../../bin/
./tensorrt_inference MiniFASNet ../configs/MiniFASNet/config.yaml ../samples/antispoofing

5.Results

name	sample	result
0.jpg		fake
1.jpg		fake
2.jpg		real
3.jpg		real
4.jpg		fake
5.jpg		fake