Question about pretrained model and results
Closed this issue · 2 comments
Ian0926 commented
Hello, thanks for your great work.
I have tried enhancing the low-light images of LOL testing set by using the provided pretrained model. The numerical results are listed as follows: (PSNR/SSIM/MAE 19.8416/0.8243/0.1166), which is inconsistent with the reported results (21.3282/0.8348/0.0832). Could provide the pretrained model used in the paper?
AndersonYong commented
Hi, noted that the illumination adjustment in our model is determined by the specific ratio. Obviously, you can carefully tune the ratio to get the best results. I've update the newest evaluation code, please use it for evaluation.
I've check the pretrained model and it worked well.
Ian0926 commented
Thanks for your response.