-
Notifications
You must be signed in to change notification settings - Fork 2.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Why paired image for sketch is ground truth in loss function? #4
Comments
V(x) is shaped as 1x1x4096 |
From my understanding, the sketch isn't necessary? If you decrease the
input to content & style, you can still get the *style transfer* done.
…On Fri, Aug 18, 2017 at 1:21 PM, lllyasviel ***@***.***> wrote:
V(x) is shaped as 1x1x4096
G(x, y) is shaped as 256x256x3 or 512x512x3
you can not make diff for them.
v(x) is style of x
G(x, y) means "use sketch of x and style of x to generate x itself."
When the dataset is large enough, it somehow works like style transfer.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#4 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGnAViVwUeDFs_RQzHAvE_dt1hFVjaO6ks5sZR9KgaJpZM4O6bvv>
.
--
Kind regards,
Dex
|
maybe and you can try |
I'm still not quite follow why you use `y`(sketch-paired image) as the
ground truth in this case? Can you explain a bit more? Thank you!
…On Fri, Aug 18, 2017 at 1:36 PM, lllyasviel ***@***.***> wrote:
maybe and you can try
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#4 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGnAVj_0PSIOz1BTQWnhagEO5vthq2uVks5sZSLYgaJpZM4O6bvv>
.
--
Kind regards,
Dex
|
Code is coming soon. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Hi!
The loss function in the paper is defined as following:
I was wondering why the
y
isthe paired domain of painting
. From my understanding, the ground truth should be style image so the output will be more close to the style you want transferred to.Does it make more sense to use like
V(x) - G(x, y)
in this case?The text was updated successfully, but these errors were encountered: