IMAGE CAPTIONING WITH DIFFUSION MODELS

Summary from the TLDR AI newsletter

This study revisits using diffusion models for image-to-text generation. It introduces the LaDiC architecture, which enhances diffusion models' performance in image captioning tasks.