some problems about "save-feature" process and "semi-supervised" process

Dayan-Guan / USRN

Code for <Unbiased Subclass Regularization for Semi-Supervised Semantic Segmentation> in CVPR 2022

27 stars 5 forks source link

some problems about "save-feature" process and "semi-supervised" process #4

Open RFKxavieryang opened 2 years ago

RFKxavieryang commented 2 years ago

Why the "save_feature" process saves the features of the val set, and then the "clustering" process also uses these features. But the "semi-supervised" process requires the clustering results of the label set. Is there something wrong?

Dayan-Guan commented 2 years ago

Hi @RFKxavieryang , thanks for your information. This issue has been corrected as: