search

yulunzhang / RCAN

PyTorch code for our ECCV 2018 paper "Image Super-Resolution Using Very Deep Residual Channel Attention Networks"
1.37k stars 313 forks source link

About preprocess of dataset #42

Open lkyee opened 5 years ago

lkyee commented 5 years ago

Hi, I want to know whether the PSNR will decrease if I only let RGB channels divide 255.0 rather than substract the mean of dataset and then add it to the output. Look forward to your kind reply！

yulunzhang commented 5 years ago

Hi I didn't compare results with and without mean subtraction. According to the experience of Lim et al., it should help stabilize the training but doesn't affect the PSNR obviously. https://github.com/thstkdgus35/EDSR-PyTorch/issues/71

lkyee commented 5 years ago

Thank you for your kind reply! I am training the code of RCAN and I changed a little.The scale I set X4.From other issues,I know you use set5 dataset as a validataion set.Can you send me the PSNR data of Set5 during the train process.I just want to compare with my model.If my model perform worse,I can stop it in time for I have no enough gpu to use.My email address is 444502842@qq.com. Look forward to your kind reply！