seresnext

SEResNeXt101 PyTorch=>ONNX=>TensorRT

1.Reference

• ResNeXt: Aggregated Residual Transformations for Deep Neural Networks
• SENet: Squeeze-and-Excitation Networks
• Pretrained models: https://data.lip6.fr/cadene/pretrainedmodels/

2.Export ONNX Model

python3 export_onnx.py

3.Build seresnext_trt Project

mkdir build && cd build
cmake ..
make -j

4.run seresnext_trt

./seresnext_trt ../config.yaml ../samples