-
-
Notifications
You must be signed in to change notification settings - Fork 1.1k
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
Allow scaling factor of 2 or 3 for REAL-ESRGAN #1313
Comments
When you're saying "previous versions" you mean <6.0.0? The reason why Real-ESRGAN supports the scales it does it limited by the models available: As you can see, we only have a 4x model for realesrgan-plus. I tried to add the new 2x model at #1289, but the model conversion doesn't work with pnnx. |
There is a Anime Plus Model which is 2x, trained by Sudo based on Anime Plus 4x. Maybe WE could use this for the Anime Plus 2x scale. |
Thanks for the response. I see your point, and yes, it was before 6 (I have 2.8.1 on my laptop and a newer one on my main PC but I can't recall the exact version, like 3 or 4?). In any case, I've no idea whether the anime ones will be good for real life videos though, but perhaps the suggestion by Pete4k may work? |
There's an official 2x model here: https://github.com/xinntao/Real-ESRGAN/releases/tag/v0.2.1 The issue is that it's in PyTorch format. I tried to convert it into ncnn which we are using, and pnnx fails to convert it (produces black frames). Someone else also tried to do the conversion with chaiNNer and it also failed. I'm not quite sure why. |
Did not someone try to send a pull request for this last Saturday? |
I don't think so. There was only a PR for RIFE |
Oh,ok |
In previous versions of Video2X, a lot drivers allowed a smaller scaling factor. With the new version, these drivers have been replaced with a choice of only 3, and the one dubbed REAL-ESRGAN, recommended for real life videos, forces a scaling factor of 4. This is a bit excessive, and smaller scaling factors would be appreciated.
The text was updated successfully, but these errors were encountered: