Mask RCNN and AffordanceNet in Pytorch

Hi, I’m wondering if anyone interested in re-implement a system like Mask RCNN in pytorch? There’re some repos in pytorch but the authors stop working on it.

Although we don’t have the source code for Mask RCNN yet but I found a new source that can do similar thing in Caffe: AffordanceNet I’m thinking about moving it to pytorch. Anyone interested?

Hey Jacob, could you complete this? I am interested in implementing it

I am also interested about that, any progress now? The AffordanceNet provide a caffe version source codes, but it does work well now, even not support cudnn7 and cuda9, very old caffe

I’m kind of late to post, but I found a few github projects on Mask RCNN in PyTorch:


@rkosti I think there already a good implementation on MaskRCNN which is maskrcnn_benchmark written by FB official.

But 2 stages are really slow, we need some 1-stage instance segmentation method implementation