search

jmiller656 / EDSR-Tensorflow

Tensorflow implementation of Enhanced Deep Residual Networks for Single Image Super-Resolution
MIT License
330 stars 107 forks source link

Did you use the GAN structure？ #51

Closed jackfrost168 closed 4 years ago

jmiller656 commented 4 years ago

No, this uses resnet-style layers and does pixel shuffling rather than convolution to upscale

jackfrost168 commented 4 years ago

Thank you！

jackfrost168 commented 4 years ago

And I have another question. The EDSR's paper said it is based on SRResNet architecture which is used gan architecture. EDSR is removing unnecessary modules. In SRResNet，it use pixel shuffling too. From my point of view，EDSR should also use the gan architecture because it is based on SRResNet as the paper said，am I right？

jmiller656 commented 4 years ago

Just because it takes inspiration from SRResNet, does not need it should use the same architecture. There isn't necessarily a "right answer" in terms of NN architectures as far as I'm aware.

It's also worth making clear: I am not the person who wrote the EDSR paper. I just happen to provide a TensorFlow implementation of their work on Github.

jackfrost168 commented 4 years ago

Thank you for your replying！