SwinIR VS SwinIR

We’re referencing our three models directly as git repos (because they’re not available as packages in PyPI) in model_requirements.txt, but we had to make a few changes to make these repos work as pip packages. Pip looks for a setup.py file in the git repo to figure out which files from the repo need to be installed into the environment, as well as what the dependencies of that repo are. GLID-3-xl and latent-diffusion (another diffusion model that GLID-3-xl depends on) had setup.py files that needed tweaks to include all of the code needed to run the models. SwinIR didn’t have a setup.py file at all, so we added one. Finally, all of these setup.py files needed additional dependencies, which we just added to the model_requirements.txt file.

003_realSR_BSRGAN_DFOWMFC_s64w8_SwinIR-L_x4_GAN.pth

I was trying to get a new upscaler working, as someone pointed me to one that did a good job of preserving and creating new details: https://github.com/JingyunLiang/SwinIR

SwinIR: https://github.com/JingyunLiang/SwinIR

Upscalers that I use SwinIR https://github.com/JingyunLiang/SwinIR https://github.com/AUTOMATIC1111/stable-diffusion-webui (Use 'extras' tab for the upscaler function) Topaz Gigapixel AI https://www.topazlabs.com/gigapixel-ai

You could try out chaiNNer, it is a free local/offline application. There are a lot of (upscaling) models which you can download an use with it. You can for example try out SwinIR-L (link will start a model download) or any other model you like depending on your input images.

the relative attention bias can be made non-square in the original implementation, there is a parameter window_size, at 7, that is forced to (7,7) directly, but you can change it easily. https://github.com/JingyunLiang/SwinIR/blob/main/models/network_swinir.py

Use SwinIR medium model to upscale by 2 times. This will result in a video of 2048x1152.