multimodallearning/pytorch-mask-rcnn

NMS without CUDA

kahya-se opened this issue · 2 comments

Hello, I'd like to use Mask-RCNN in a CPU only mode. Yet still, the NMS module has to be installed using CUDA, which seems to me unnecessary, as I don't intend to use a GPU [for testing purposes on this machine]. Isn't it possible to run the model without CUDA (as it is possible in the tf-mask-rcnn)?

Duplicate to issue #15

Have you solved the problem yet?