[REQ] SCUNet and colab

[MarcoRavich]

Hi there, 1st of all thanks for your (voluntary) work !

It would be cool to have the "Practical Blind Denoising via Swin-Conv-UNet and Data Synthesis" implemented in aydin:

git: https://github.com/cszn/SCUNet online demo: https://replicate.com/cszn/scunet vs-implementarion: https://github.com/HolyWu/vs-scunet

Last but not least it would be also cool to "transpose" aydin into a Jupiter notebook like these: https://colab.research.google.com/github/jantic/DeOldify/blob/master/ImageColorizerColabStable.ipynb https://github.com/XPixelGroup/BasicSR/tree/master/colab https://replicate.com/cszn/scunet ...

Hope that inspires !

note: just pushed aydin @ Doom9 too

[AhmetCanSolak]

Hello @forart ! Thank you for creating a feature request.

Publication you suggested(SCUNet) looks interesting and I will be looking into its details(both the paper and their implementation). I have quite a lot on my plate these days but I will try to give you an update on this next week.

PS: Thanks for mentioning Aydin in other forums and I want to clarify that Aydin works on spatio-temporal data(TZYX) when all four dimensions are available and can do TYX as well. Aydin doesn't read video file formats directly though, if you can save it as a file format that is supported then you can denoise movies too. With biological live imaging datasets, we use Aydin for denoising movies on daily basis. I am clarifying this following another person's comment here: https://forum.doom9.org/showthread.php?p=1982241#post1982241

[MarcoRavich]

1st of all, thanks for your fast reply.

And, of course, take your time to check @cszn's repos (where you can find other interesting denoisers - and not only - too): https://github.com/cszn?tab=repositories

Last but not least, since aydin can work on image sequences (= videos), we do also STRONGLY suggest you to open a account @ Doom9 to attract more users and - why not - collaborators.

Thanks again.

[AhmetCanSolak]

Hello @forart ,

I read the SCUNet paper and it is highly relevant for our package. We can potentially have it as one of our advanced mode algorithms (I noticed we don't talk on this in GUI tutorials in docs and adding it here: https://github.com/royerlab/aydin/pull/294). I will investigate on a potential torch implementation and make a PR accordingly.

Currently, I do not have bandwidth to engage with users on a new platform but both you and other Doom9 users welcome to make bug reports/feature requests here on GitHub.

[MarcoRavich]

Well, glad to be useful for the project...

...about the implementation of this and other algos, I believe that @Selur can suggest you the best resources to achieve efficently.

Hope that helps.