AISP

Why we included this project

Low-level computer vision is the part of the field that deals with the camera pipeline itself, and this repository gathers deep learning methods for that territory: RAW photo processing, denoising, deblurring, super-resolution, and learned image signal processing. Teams working in those areas will find working code and training tutorials, including a notebook that walks through generating realistic degraded RAW images. It also bundles several peer-reviewed models, such as a lightweight network for perceptual image enhancement that can run on smartphones and a bokeh rendering approach, so it works both as a reference for reproducing published results and as a starting point for adapting these techniques to real camera products. Since it is the official codebase for NTIRE and AIM challenges, it is a good place to see how current restoration methods are structured and evaluated.