The result of given pretrained model WGSR(2lvl)

[WeiYangHui]

Dear author, I tested the DIV2K dataset using the model WGSR (2-lvl) you provided, and the results are as follows:


May I ask why PSNR and SSIM differ so much from the results you provided in the article?May I ask why PSNR and SSIM differ so much from the results you provided in the article?

[mandalinadagi]

Sure, as convention in SR evaluation, we calculated PSNR and SSIM values on Y-channel images not from RGB ones. Calculation of PSNR values either on RGB or Y channel yield almost 2 dB difference. For further discussion you can refer to https://arxiv.org/pdf/2104.14868.pdf

Also as requested, DIV2K results are added to Google Drive, you can download from here: https://drive.google.com/drive/folders/19Ur1SF-49erDA8fOoZAx1_Pnt7ZKRT0X

[WeiYangHui]

Thanks very much.