Open lanbowan1 opened 3 months ago
Apologies for the delayed response; my old email address linked to GitHub has expired.
PSNR and SSIM values vary significantly depending on the implementation and the version of the package used. Unfortunately, I cannot recall the exact package and version I used, as it has been three years.
However, you can compare the visual results with those presented in my paper. If you notice any inconsistencies, please let me know, and I will double-check the checkpoints. Besides, you can show me your implementations of PSNR and SSIM, and I can look at it.
Hi, I have tested and used the pre-training weights file(weight/Epoch99.pth) you gave and added PSNR with SSIM calculations and found that the values are very different, can you please help me to analyse the reason? psnr(10.60)and ssim(0.36)