about the loss - Githubissues

Elegantice commented 4 years ago

sorry, maybe i dont read your article in detail. i cant find the introduce about the loss.And your code has much loss. i think they are the different loss of different model. can u give me a introduce for your loss? thank u very much.

Elegantice commented 4 years ago

In detail,i can not find the out, out2, out 3, out 4.just can find the sr.so idont have the complete loss.because i cant get the loss1, loss2,loss3. The reason is that I can only get one result from the self.model(lr, idx_scale).look forward to your reply.

Feynman1999 commented 4 years ago

@Elegantice I also don't know. I guss L1 loss? but i can't reproduce the results in paper...

Sudo-Biao commented 4 years ago

L 1 loss, and you can check with the pre-trained model , or send me your problem.发自我的华为手机-------- 原始邮件 --------发件人： Yuxiang Chen notifications@github.com日期： 2020年1月6日周一傍晚5:27收件人： Sudo-Biao/s-LWSR s-LWSR@noreply.github.com抄送： Subscribed subscribed@noreply.github.com主题： Re: [Sudo-Biao/s-LWSR] about the loss (#3)@Elegantice I also don't know. I guss L1 loss? but i can't reproduce the results in paper...

—You are receiving this because you are subscribed to this thread.Reply to this email directly, view it on GitHub, or unsubscribe.

Feynman1999 commented 4 years ago

thanks for your reply! I have some questions below: *I found in your code that the input images is not divided by 255. Is there any difference between this and dividing by 255 to map to 0 ~ 1? I think there is no difference, so my code is mapped to 0 ~ 1.

What has data enhancement done specifically? How much do these actions affect the final result?
The value range of the output image is not exactly 0-255, is it 0 if it is less than 0, or 255 If it is greater than 255 (clamp operation)?
How many epochs are trained by the results shown in the paper?

Thank you for taking the time to read and give advice.

yuxiang chen

Biao notifications@github.com 于2020年1月6日周一下午5:44写道：

L 1 loss, and you can check with the pre-trained model , or send me your problem.发自我的华为手机-------- 原始邮件 --------发件人： Yuxiang Chen < notifications@github.com>日期： 2020年1月6日周一傍晚5:27收件人： Sudo-Biao/s-LWSR < s-LWSR@noreply.github.com>抄送： Subscribed subscribed@noreply.github.com主题： Re: [Sudo-Biao/s-LWSR] about the loss (#3)@Elegantice I also don't know. I guss L1 loss? but i can't reproduce the results in paper...

—You are receiving this because you are subscribed to this thread.Reply to this email directly, view it on GitHub, or unsubscribe.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Sudo-Biao/s-LWSR/issues/3?email_source=notifications&email_token=AGC5WRF5G3ZSDMBYWEHJM2DQ4L4RJA5CNFSM4J6U3QV2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEIE6DLY#issuecomment-571072943, or unsubscribe https://github.com/notifications/unsubscribe-auth/AGC5WRD27XCLO6IFDU7DKPTQ4L4RJANCNFSM4J6U3QVQ .

Sudo-Biao commented 4 years ago

sorry, maybe i dont read your article in detail. i cant find the introduce about the loss.And your code has much loss. i think they are the different loss of different model. can u give me a introduce for your loss? thank u very much.

Sorry for the mistake. I update the right code now.

Sudo-Biao commented 4 years ago

@Elegantice I also don't know. I guss L1 loss? but i can't reproduce the results in paper...

Sorry for the mistake, and I have updated the right code. You can try it.

Sudo-Biao / s-LWSR

about the loss #3