search

chaiNNer-org / spandrel

Spandrel gives your project support for various PyTorch architectures meant for AI Super-Resolution, restoration, and inpainting. Based on the model support implemented in chaiNNer.
MIT License
155 stars 12 forks source link

[Feature Request] Denoising, deblurring, JPEG artefact removal, and colourisation #16

Closed Wurlit closed 11 months ago

Wurlit commented 1 year ago

After looking at various image correction methods, and since Chainner already works for upscaling and face restoration models, I'd like to request support for denoising, deblurring, JPEG artefact removal, and colourisation at some point if possible.

An example model for each method:

Denoising: SCUNet https://phhofm.github.io/upscale/assets/denoise_example.d58a8997.jpg

Deblurring: MAXIM Go Pro https://phhofm.github.io/upscale/assets/deblur_example.03ca624e.jpg

JPEG artefact removal: FBCNN (fbcnn_color.pth) https://phhofm.github.io/upscale/assets/tomandjerry.f662f514.png

Colourisation: Codeformer (codeformer_colorization.pth) https://github.com/sczhou/CodeFormer/blob/master/assets/color_enhancement_result1.png

I haven't been able to find them all as .pth models though.

Stl5n0 commented 1 year ago

I was curious about a few different models and found this- https://github.com/PINTO0309/PINTO_model_zoo/blob/main/README.md Tons of different models available as .onnx, including FBCNN 😉

RunDevelopment commented 11 months ago

Deblurring: MAXIM Go Pro

This isn't a PyTorch model, so I don't think we can add support for it.

All other architectures are supported now, so I'll close this issue.