- TensorRT 7.2.3.4
- gcc 7.5.0
- cmake 3.10.2
- CUDA 11.1.1
- cuDNN 8.1.1
- openCV 4.5.1
- change
DOpenCV_DIR
andDTensorRT_DIR
to your OpenCV and TensorRT installation inbuild.sh
and runsh build.sh
(requires exporting PyTorch model to ONNX model first, this repo only converts models from ONNX to TensorRT)
- run
./bin/serialize
to build tensorRT engine from ONNX model and serialize it - run
./bin/inference
to run inference