search

HiLab-git / WSL4MIS

Scribbles or Points-based weakly-supervised learning for medical image segmentation, a strong baseline, and tutorial for research and application.
MIT License
208 stars 34 forks source link

quesition on codes #1

Closed wwjwy closed 3 years ago

wwjwy commented 3 years ago

Sorry to interrupt you, I noticed that your weakly supervised and semi-supervised codes seem to differ only in the loss function. Therefore, I think that the semi-supervised model can be modified using a loss function to get a simple weakly supervised model. Can i understand this way? Thanks

Luoxd1996 commented 3 years ago

Hi, Compared with fully-supervised learning, most of the semi/weakly-supervised learning just used different training strategies to learn from imperfect annotation. So I think your understanding is right. Best, Xiangde.