About rectification in stage 2

layumi / Seg-Uncertainty

IJCAI2020 & IJCV2021 :city_sunrise: Unsupervised Scene Adaptation with Memory Regularization in vivo

https://arxiv.org/abs/1912.11164

MIT License

387 stars 51 forks source link

About rectification in stage 2 #6

Closed adricarda closed 4 years ago

adricarda commented 4 years ago

Hi, thanks for the code! I was wondering why in stage 2 the loss weight lambda-kl-target is set to 0. From my understanding, the rectification is done by using this term. Thanks in advance.

adricarda commented 4 years ago

The lambda-kl-target term is 0 because it is only used in stage 1 when minimizing the discrepancy between the two classifiers.