search

cchen156 / Learning-to-See-in-the-Dark

Learning to See in the Dark. CVPR 2018
http://cchen156.web.engr.illinois.edu/SID.html
MIT License
5.45k stars 842 forks source link

Questions regarding to the PSNR/SSIM results #23

Open wbhu opened 6 years ago

wbhu commented 6 years ago

Hi @cchen156 ,

This work is very exciting. But I have a question about the quantitative results reported in the paper Table3.

I downloaded the source code and trained model, and tested Sony dataset. I generate GT by Rawpy and apply Gamma transfer after it, since you apply Gamma transfer on the output of the network in your test code. I found that the results looks fine, but the average PSNR/SSIM in the testset is 25.44/0.713, which is quite lower than 28.88/0.787 reported in Table3 in your paper.

If I have made some mistake? How can I get the reported results.

Thanks, Wenbo

cchen156 commented 6 years ago

Hi, rawpy already applied gamma correction by default. I think you applied gamma correction twice in this case.

cchen156 commented 6 years ago

The rawpy. postprocess contains gamma correction.

How did you train it with 1 hour and get 30.2dB psnr. This is interesting!

wbhu commented 6 years ago

Could you please share your trained model @eedalong . Thanks a lot.

xiaojidan commented 6 years ago

@eedalong can you share your code?

KeqiWangSXuniversity commented 6 years ago

hello， @cchen156 I also encountered such a problem. The average PSNR/SSIM in the Sony testset is 28.64/0.7674., which is lower than 28.88/0.787 reported in Table3 in your paper.I downloaded the model you provided, but the results were very poor. I felt that the model could not be loaded correctly, but I checked my path and there was no problem.As with the code provided by your github, no changes have been made.English is not very good, please forgive me.

KeqiWangSXuniversity commented 6 years ago

Hello, I downloaded the ckpt you only have two files, model.ckpt.data-00000-of-00001, model.ckpt.meta, is my download error, or what caused it? Should there be four files, and should also include checkpoint, model.ckpt.index. Due to the lack of these two files, this model is difficult to load into, is it really a problem with my operation?

cchen156 commented 6 years ago

Hi can you try the Baidu link: https://pan.baidu.com/s/1WfijVSQfBimQLCU_QL_Iaw

KeqiWangSXuniversity commented 6 years ago

Hi can you try the Baidu link: https://pan.baidu.com/s/1WfijVSQfBimQLCU_QL_Iaw

Ok, thank you

lyn-max commented 5 years ago

@KeqiWangSXuniversity Hi, have you solved the pretrained-model problem?