You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, no error was reported when I used the ssim function api to calculate the ssim value, but the calculated ssim value was negative. This is not normal, because the ssim value should be [0-1], but I don't know what is wrong, do you have any suggestions?
The text was updated successfully, but these errors were encountered:
@xiaoyangx0 could you provide a minimal example of the code + data that results in negative values? We do not observe such behavior on data from general IQA datasets e.g., TID2013, KADID-10k and PIPAL.
shadow_ssim += ssim(original_data, shadow_inv_output_, reduction='sum').item() shadow_psnr += psnr(original_data, shadow_inv_output_, reduction='sum').item()
Thank you for your reply. My original intention was to generate a picture through a model and compare the similarity with the original picture. The code for using the ssim has been uploaded. However, no bug was displayed when I used the ssim. However, when I used tensorboard to display the ssim change curve, I found that some places were negative. <img width="479" alt="image" src="https://user-images.githubusercontent.com/62381232/223094975-2b5f2603-6004-4f16-a5f6-3056c4d9170e.png">
Hello, no error was reported when I used the ssim function api to calculate the ssim value, but the calculated ssim value was negative. This is not normal, because the ssim value should be [0-1], but I don't know what is wrong, do you have any suggestions?
The text was updated successfully, but these errors were encountered: