KAIR (Kai’s AI Image Restoration) provides implementation and training codes for a collection of deep learning models used in computer vision image restoration tasks. The toolbox covers denoising models (DnCNN, FFDNet, SCUNet), super-resolution models (ESRGAN, SwinIR, MSRResNet), and video restoration models (VRT, RVRT). It includes utilities for model training, testing, and FLOP estimation across these computer vision tasks.