Closed dlongry closed 7 years ago
Follow the commented out procedure. First pretraijed the Discriminator then the SRGAN. The pre training for SRGAN was a test to see if it was learning anything on its own or not.
I don't really think this works to be honest. If you are hoping to get a upscaled image then i suggest going for the theano implementation by jcjohnson.
Somshubra Majumdar
On Mar 31, 2017 21:32, "dlongry" notifications@github.com wrote:
Thank you for this great work. but I have a doubt when I using your model.py to train: Should I pretrain the SR network at first,and then pretrain the Discriminator network.at last train the full network?
or just directly train the full network?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/titu1994/Super-Resolution-using-Generative-Adversarial-Networks/issues/22, or mute the thread https://github.com/notifications/unsubscribe-auth/AC6Emr2flUybrTGbcoem2U6wX0ZwVY2Fks5rrbc6gaJpZM4MwVMR .
Thanks a lot. And could you give me the link to SRGAN implementation by jcjohnsom? @titu1994
Just search in Google
Somshubra Majumdar
On Apr 3, 2017 20:54, "dlongry" notifications@github.com wrote:
Thanks a lot. And could you give me the link to SRGAN implementation by jcjohnsom?
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/titu1994/Super-Resolution-using-Generative-Adversarial-Networks/issues/22#issuecomment-291370535, or mute the thread https://github.com/notifications/unsubscribe-auth/AC6Emna_4C1UCspCsgUat2hLmAXFs4wOks5rsaLtgaJpZM4MwVMR .
Thanks for quickly reply. In fact,I have searched it in Google before I update my quetion. Maybe I used the fake Google. e... I can't find the SRGAN implementation by jcjohnsom that you said. so...Could you give me a hand?Thanks.@titu1994
Sorry I was busy with an assignment deadline at that time. I made a mistake with the name. Its https://github.com/alexjc/neural-enhance.
Alexjc rather than jcjohnson. Sorry for the confusion.
I get it. Thank you a lot for replying to me from your busy schedule.
Thank you for this great work. but I have a doubt when I using your model.py to train: Should I pretrain the SR network at first,then pretrain the Discriminator network. And at last, train the full network?
or just directly train the full network?
@titu1994