search

sanghyun-son / EDSR-PyTorch

PyTorch version of the paper 'Enhanced Deep Residual Networks for Single Image Super-Resolution' (CVPRW 2017)
MIT License
2.41k stars 666 forks source link

The code of the RDN model does not match the code of the original paper #305

Open HWalkingMan opened 3 years ago

HWalkingMan commented 3 years ago

In EDSR-Pytorch: https://github.com/sanghyun-son/EDSR-PyTorch/blob/9d3bb0ec620ea2ac1b5e5e7a32b0133fbba66fd2/src/model/rdn.py#L88

In Original Paper Repo: https://github.com/yulunzhang/RDN/blob/fdc932a815976f7546416b7cca4b3f016034e2cd/RDN_TrainCode/code/model/RDN.lua#L126

the input_channel of the last Conv layer in UpNet should be G0 instead of G, beacuse the input_channel of original paper repo used opt.nFeat instead of opt.growthRate

ryanxingql commented 1 year ago

I have the same issue.