My question is based on the paper. This paper introduced a loss function by calculating two-loss: the loss between the white noise input and style image and the loss between the white noise input and content image. So my question is, why is the loss function not include the loss between the style image and content image?


  • $\begingroup$ How about adding a link to the paper ? Otherwise we really don't have much of an idea what you are talking about. $\endgroup$
    – Hilmar
    Nov 26, 2022 at 13:33
  • $\begingroup$ @Hilmar arxiv.org/abs/1508.06576 $\endgroup$
    – sss
    Nov 26, 2022 at 14:22
  • $\begingroup$ Opinion: that would be a very good question to ask the authors, at the email address they included. I don't think there's a good answer to your "why", but based on the fact that the authors seem to want to take the style of some painting and then apply it to photographs, I suspect they felt that comparing the style that they've abstracted out of a painting with the content that they had abstracted out of a photograph would not be very meaningful. $\endgroup$
    – TimWescott
    Nov 26, 2022 at 22:37
  • $\begingroup$ @TimWescott That'll be a good point of why $\endgroup$
    – sss
    Nov 27, 2022 at 9:46
  • $\begingroup$ Well, yes, but at least this stackexchange discourages questions that solicit opinions, and answers that are opinions. The real "why" is inside the head of the authors, which is why I suggested writing to them (then, maybe, answering your own question here, because it is an interesting one). $\endgroup$
    – TimWescott
    Nov 27, 2022 at 19:16


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.