-
Notifications
You must be signed in to change notification settings - Fork 247
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
EXIF rotation and accessible colab for image completion with more features #82
Comments
Really cool stuff! Outpainting was the one thing I really missed in the original implementation, is it possible to add a custom mask instead of just cropping the image? Or is there a technical limitation that prevents that? |
the mask works on the 32x32 representation so this is one limit. also the encoding is sequential afaiu from top to bottom, so completion is mostly relevant downwards. it would have been awesome if we could do completion for specific color (e.g. all white pixels). but i dont think this could be done here. maybe the custom resolution encoding can help? didn't get to looking into that yet. |
hi, like others, i prepared a colab with more accessible GUI and presets here:
https://gist.github.com/eyaler/0cee9a71f5dd3fdfa9c0c03656ebdd4c
this is oriented for easily doing batch experiments and you can see some results here:
https://twitter.com/eyaler/status/1468682110860992521
there are some very simple things you may want to take from the notebook as fixing image rotation due to EXIF:
im = ImageOps.exif_transpose(Image.open(file))
and fixing the aspect ratio in post-process back to the original (although it would be better to use non-square encoding).
thanks fur rudalle!
The text was updated successfully, but these errors were encountered: