A great write up, and don’t miss the publications section of his website. The paper on multiframe superresolution is a tour-de-force and crushes a ton of deep learning results despite not being learning based.