HiDiffusion: Unlocking Higher-Resolution Creativity and Efficiency in Pretrained Diffusion Models

TL;DR

HiDiffusion is a tuning-free framework that enhances high-resolution image synthesis in pretrained diffusion models by dynamically adjusting features and reducing self-attention redundancy, leading to faster and more accurate results.

Contribution

The paper introduces HiDiffusion, a novel framework with Resolution-Aware U-Net and optimized self-attention, enabling higher-resolution image generation without additional training.

Findings

Supports resolutions up to 4096x4096

Achieves 1.5-6x inference speedup

Outperforms previous methods in quality and efficiency

Abstract

Diffusion models have become a mainstream approach for high-resolution image synthesis. However, directly generating higher-resolution images from pretrained diffusion models will encounter unreasonable object duplication and exponentially increase the generation time. In this paper, we discover that object duplication arises from feature duplication in the deep blocks of the U-Net. Concurrently, We pinpoint the extended generation times to self-attention redundancy in U-Net's top blocks. To address these issues, we propose a tuning-free higher-resolution framework named HiDiffusion. Specifically, HiDiffusion contains Resolution-Aware U-Net (RAU-Net) that dynamically adjusts the feature map size to resolve object duplication and engages Modified Shifted Window Multi-head Self-Attention (MSW-MSA) that utilizes optimized window attention to reduce computations. we can integrate HiDiffusion into various pretrained diffusion models to scale image generation resolutions even to 4096x4096 at 1.5-6x the inference speed of previous methods. Extensive experiments demonstrate that our approach can address object duplication and heavy computation issues, achieving state-of-the-art performance on higher-resolution image synthesis tasks.