search

Algolzw / image-restoration-sde

Image Restoration with Mean-Reverting Stochastic Differential Equations, ICML 2023. Winning solution of the NTIRE 2023 Image Shadow Removal Challenge.
https://algolzw.github.io/ir-sde/index.html
MIT License
524 stars 38 forks source link

test.py #8

Open treefreq opened 1 year ago

treefreq commented 1 year ago

Why does GT data need to be used for testing in the deraining/test.py ? In my understanding, during testing, only LQ data is needed, and GT data cannot be involved. Maybe I didn't understand

image

Algolzw commented 1 year ago

Hi! Actually we never use it in model testing (only for computing PSNR/SSIM/LPIPS). It is ok to remove it from the testing code.

treefreq commented 1 year ago

Thank you for your reply

I see that the 'GT' parameter can be 'none', and I will remove it to run

flynnamy commented 1 year ago

Hi,though it can be none,it was trained by use LQ and the part of GT. If I only use LQ to do inference,the results are not good. How to solve this?@Algolzw @treefreq

treefreq commented 1 year ago

GT data is not involved in the testing process, it is only used for PSNR calculation during testing; Therefore, even if you add GT data, I believe your results will remain unchanged. Something went wrong with your training or elsewhere