This research introduces SVG, a novel latent diffusion framework that eliminates variational autoencoders by leveraging frozen DINO features for structured l...
Level: advanced
By Minglei Shi and 8 other authors
Category: education