Generative Refinement Networks for Visual Synthesis

Jian Han, Jinlai Liu, Jiahuan Wang +2

While diffusion models dominate the field of visual generation, they are computationally inefficient, applying a uniform computational effort regardless of different complexity. In…