Not All Steps are Created Equal: Selective Diffusion Distillation for Image Manipulation

Abstract

Conditional diffusion models have demonstrated impressive performance in image manipulation tasks. The general pipeline involves adding noise to the image and then denoising it. However, this method faces a trade-off problem: adding too much noise affects the fidelity of the image while adding too little affects its editability. This largely limits their practical applicability. In this paper, we propose a novel framework, Selective Diffusion Distillation (SDD), that ensures both the fidelity and editability of images. Instead of directly editing images with a diffusion model, we train a feedforward image manipulation network under the guidance of the diffusion model. Besides, we propose an effective indicator to select the semantic-related timestep to obtain the correct semantic guidance from the diffusion model. This approach successfully avoids the dilemma caused by the diffusion process. Our extensive experiments demonstrate the advantages of our framework.

Ying-Cong Chen

Assistant Professor

Ying-Cong Chen is an Assistant Professor at AI Thrust, Information Hub of Hong Kong University of Science and Technology (Guangzhou Campus). He obtained his Ph.D. degree from the Chinese University of Hong Kong. His research lies in the broad area of computer vision and machine learning, aiming for empowering machine with the capacity to understand human appearance, physiology and psychology. His works contribute to a wide range of applications, including contactless health monitoring, semantic photo synthesis, and intelligent video surveillance.