search

iSEE-Laboratory / DiffUIR

The official implementation of the paper of CVPR2024: Selective Hourglass Mapping for Universal Image Restoration Based on Diffusion Model
MIT License
103 stars 7 forks source link

LOL dataset results were not noticeable #14

Closed lanpokn closed 2 months ago

lanpokn commented 2 months ago

I tried to test the low-light enhancement task on the LOL dataset with visualize.py, and the results were not noticeable, and the image was basically unchanged, why is that? (I have tested Gopro dataset and it is good) before: 112 after: temp1

zhengdian1 commented 2 months ago

Note that if you want to test low-light image, please use the code src/visualization-Line1279-1281

lanpokn commented 2 months ago

lines 1279-1281: “self.ema.to(self.device) print("test start") if self.condition”

It seems that as long as visual.py is called, what is actually invoked is src/visualization at lines 1279-1281 . Therefore, it appears that calling these lines directly may not resolve the issue.

zhengdian1 commented 2 months ago

The code is modified one times, if you git recently, it should be lines 1286-1288