About experiment results

[liuxuuuu]

Hello, I use SIDD-Medium dataset and get 24,542 cropped images after prepare and changed in the config file by replacing '64' to 'None' for 'n_data', or removing the highlighted line. (below image).

Then I run python train.py -c APBSN_SIDD -g 0 .
But the performance seems not as good as that mentioned in the paper.

I'm confused and want to confirm the experiment setting. Thanks!

[wooseoklee4]

Thanks for your interest in our work!

Could you please tell me which performance is worse than the number in paper?
If the SSIM is matter, you may check the exact the hyperparameter of SSIM measurement. (like data range)

Also note that, there is a difference between valiation dataset and test(benchmark) dataset.

Thanks

[liuxuuuu]

Thank you for your reply! The result in the above figure is the validation dataset I used for training. My question is that SSIM did not perform as well as in the paper. Could you tell me where to modify the hyperparameter of SSIM measurement? Thank you very much