Skip to content
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

real image mix #5

Closed
rainsoulsrx opened this issue Jan 7, 2020 · 4 comments
Closed

real image mix #5

rainsoulsrx opened this issue Jan 7, 2020 · 4 comments

Comments

@rainsoulsrx
Copy link

rainsoulsrx commented Jan 7, 2020

hi, thank you for your work, I have a question when doing style mixing between two real images,
In stylgan1, I chose range(8,18), which performs well, but in stylegan2, I use the same param, the results turn out to be bad.
I wonder, the scale in stylegan 1 and 2 is represent different meaning?

@dimitry12
Copy link

@rainsoulsrx I suspect, semantic content of style-layers changed in StyleGAN2. Were you able to find the range which works for StyleGAN2?

@b4nn3d
Copy link

b4nn3d commented Jan 12, 2020

can you share an how to for create and mix images from npy files?

@rolux
Copy link
Owner

rolux commented Jan 14, 2020

Style mixing works as demonstrated in the original StyleGAN2 repo. Nothing has changed there. See this comment for orientation.

If, however, you're mixing latents from encoder/projector results, this may not work as expected. In case the distance of your latents to w_avg becomes too large, they lose their "semantic content" altogether. For more, read the thread linked above.

@rolux rolux closed this as completed Jan 14, 2020
@shahik
Copy link

shahik commented Feb 4, 2020

@rainsoulsrx Plz share the code for npy files style-mixing ?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants