Writing portions of Faster RCNN in C++

Is it possible to implement certain portions of Faster RCNN in C++ ? I was thinking of things such as Non-Maximal Suppression and ROI pooling layer. I unfortunately have been unable to find any references on how to interface PyTorch with C++ for a task such as above.
But then I am very new to this library.

Thiss is an implementation of faster rcnn using pytorch. He wrote C++(CUDA) code for roi-pooling and non maximal suppression(NMS).


Pytorch docs for extending torch.nn, torch.autograd, and writing custom C extensions utilizing our C libraries.

Note: If you figure out how to write cuda code and interface it with pytorch, please tell me.

1 Like