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

TL;DR

This paper introduces Selective Diffusion Distillation, a framework that trains a feedforward network guided by diffusion models to improve image manipulation fidelity and editability, overcoming the noise trade-off problem.

Contribution

The paper proposes a novel selective diffusion distillation method with an indicator for semantic timestep selection, enhancing image manipulation quality.

Findings

Improved image fidelity and editability in manipulation tasks

Effective semantic guidance from diffusion models

Outperforms existing diffusion-based methods

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. Code is released at https://github.com/AndysonYs/Selective-Diffusion-Distillation.