Problem with training in source domain

cchen-cc / SFDA-DPL

[MICCAI'21] Source-Free domain adaptive fundus image segmentation with denoised pseudo-labeling

51 stars 9 forks source link

Problem with training in source domain #4

Open zheangh opened 1 year ago

zheangh commented 1 year ago

In train_source.py, the validation set is the dataset of target domain (see attached figures). Shouldn't you use data in source domain instead? I am sorry if I miss anything.