Utilizing Mask R-CNN without CUDA


I’m running Arch on my Surface Pro 3 and I want to use Mask R-CNN to train a detector but I don’t have NVIDIA graphics on my laptop so I can’t use CUDA. Is there a way for me to use Mask R-CNN without CUDA?


For PyTorch implementations, there’s detectron, detectron2 and the torchvision one. Based on my experience on installing and using all of them, I’d recommend the one in torchvision if you don’t use CUDA. It’s part of the torchvision package so it’s easy to get and it’s relatively easy to use.

I’m pretty sure you can use it if you have just a CPU. detectron and detectron2 require some packages which I assume require CUDA.

Do you happen to know how I can go about installing them on Pycharm?

Nvm. I’ve figured it out. Thank you very much.

1 Like