DCNv2 latest

original project: https://github.com/jinfagang/DCNv2_latest

Since DCN is used in many models and performance well but in industry this op support is not very well. Including pytorch, onnx, tensorrt etc. This repo is make DCNv2 available at all versions in pytorch.

Pytorch 1.7 inferenced in CenterNet-DLA model. It works on Pytorch 1.7 so that you can use it in your RTX 3070 series cards.

install

pytorch

./make.sh
python testcpu.py
python testgpu.py

onnxruntime

export onnx module

python onnx_test.py

install onnxruntime c++ lib

install NuGet

install onnxruntime

nuget install Microsoft.ML.OnnxRuntime -Version 1.7.0

build deconv onnxruntime

Download Eigen library, add path to CMakeLists.txt, then build.

cd ort_custom_op
mkdir build
cd build
cmake ..
make
# test
./customop

complare result to the output of onnx_test.py.

Updates

2021.02.18: Happy new year! PyTorch 1.7 finally supported on master branch! for lower version theoretically also works, if not, pls fire an issue to me!.
2020.09.23: Now master branch works for pytorch 1.6 by default, for older version you gonna need separated one.
2020.08.25: Check out pytorch1.6 branch for pytorch 1.6 support, you will meet an error like THCudaBlas_Sgemv undefined if you using pytorch 1.6 build master branch. master branch now work for pytorch 1.5;