BasicSR is an open source image and video restoration toolbox that provides implementations of various deep learning architectures for tasks including super-resolution, denoising, and deblurring. It includes models such as EDSR, RCAN, SRGAN, ESRGAN, EDVR, BasicVSR, SwinIR, and supports StyleGAN2 for generation tasks. The toolbox offers training scripts, inference pipelines, and model动物园 for computer vision researchers and practitioners.