Efficient and Training-Free Single-Image Diffusion Models

The paper presents an efficient and training-free single-image diffusion model that achieves state-of-the-art generation quality and diversity.