search

JiangtaoNie / UAL

Unsupervised Adaptation Learning for Hyperspectral Imagery Super-resolution
33 stars 11 forks source link

Unsupervised #3

Closed ZhuangChen25674 closed 2 years ago

ZhuangChen25674 commented 3 years ago

Hi, My name is ZhuangChen from HUST, I am reading your paper named "Unsupervised Adaptation Learning for Hyperspectral Imagery Super-resolution" and getting a lots of knowledge. Really thk U! While I have a question about your paper .

U say "a pre-trained fusion module whose weights are fixed in unsupervised learning" in section 4.2. But when I am reading your code in "Train_FusionModel.py" I find that U still use loss function to compute loss and use optimizer to update parameters as other supervised algorithm.

I feel so confused about that. Can u help me work this question out ? Thks!

JiangtaoNie commented 3 years ago

The "Train_FusionModel.py" is the code to pre-train the fusion model, we train it to exploit the general priors that contain in external datasets in a supervised manner. The training of adapter network is totally unsupervised.