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.

Sign up for GitHub

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

Jump to bottom

Next steps to improve SPADE + AdaIN architecture #126

Open

4 of 6 tasks

51N84D opened this issue Apr 20, 2020 · 0 comments

Open

4 of 6 tasks

Next steps to improve SPADE + AdaIN architecture #126

51N84D opened this issue Apr 20, 2020 · 0 comments

Assignees

Labels

domain:ml priority:high

Member

51N84D commented Apr 20, 2020 •

edited

Loading

We've converged to a SPADE + MUNIT hybrid, but we need some tricks to improve. Here are the short term ideas to try:

Gram Matrix as measure of style
Feature matching loss on global discriminator
Give style information directly to spade (via addition, for example)
Use decaying non-binary masks for smoothing
Use image of only water for perceptual loss
Compute perceptual loss on intersection of ground + flood mask

51N84D added priority:high domain:ml labels

51N84D self-assigned this

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment