沈云航 Yunhang Shen
沈云航 Yunhang Shen
Thanks. We are re-organizing our code in PyTorch. And we are expecting code release in this month.
Hi, the code is already available in this repository. https://github.com/shenyunhang/UWSOD/tree/UWSOD/projects/WSL
Sorry for my late response. It is complete codes. SSOPG is implemented at [here](https://github.com/shenyunhang/UWSOD/blob/UWSOD/projects/WSL/configs/PascalVOC-Detection/uwsod_WSR_18_DC5_1x.yaml#L19). SSOPG is related at [here](https://github.com/shenyunhang/UWSOD/blob/UWSOD/projects/WSL/configs/PascalVOC-Detection/uwsod_WSR_18_DC5_1x.yaml#L79).
Code is released in [CSC](https://github.com/shenyunhang/CSC.git). Please refer to the [README](https://github.com/shenyunhang/GAL-fWSD/blob/master/README.md) for details.
It seem the codes are easy to run as detectron2. And all methods only need 1-3 days to get the results, so I did not upload the models.
MEAN_LOSS is designed to norm loss over classes. For wsddn and contextlocnet, we use 128 batch size, and do not norm loss over classes, i.e., MEAN_LOSS = False. For other...
Although I do not test it on pytorch1.7 with cuda11.0, I think it should work. If you have any problem please feel free to open issues.
I just have installed it with PyTorch 1.7. Can you post the installing log?
I have re-generated the mcg proposals and did not find this problem. You may need to check the proposal file. It should contain the following keys: ``` dict_keys(['__header__', '__version__', '__globals__',...
I did not try the unsupervised or self-supervised backbones. It could be an interesting work.