Abdullah-eng40/AESRGAN ? reverse-engineered prompt

Build me this super resolution app that can take a low quality image and upscale it with a nice before and after preview in the browser.

I want a dark modern web studio where I can upload an image, zoom and pan it, compare the original and the enhanced version side by side with a slider, and see basic processing stats while it runs. Please include a backend that can handle image uploads, send the work through a fast API layer, and keep the admin side separate so I can monitor jobs and images. It should support breaking large images into tiles so huge photos do not run out of memory, then stitch them back together cleanly without visible seams.

Also set up the model inference and training pieces so the project can upscale images using the ASRGAN style pipeline already described in the repo, and make sure the whole thing is usable locally with a simple start up flow. If you need to check current library docs online, go ahead.