BETTER IMAGE RESTORATION WITH VISION-LANGUAGE MODELS

Summary from the TLDR AI newsletter

Vision-language models like CLIP work well for many tasks but struggle with image restoration. This project introduces DA-CLIP, a model that adjusts how CLIP works to handle these tasks better, leading to clearer image reconstructions.