Simple diffusion: End-to-end diffusion for high resolution images

TL;DR

This paper introduces a simplified approach to train standard diffusion models directly on high-resolution images, achieving state-of-the-art results without complex multi-stage or latent diffusion methods.

Contribution

It demonstrates that with specific adjustments like noise schedule tuning, selective scaling, dropout placement, and downsampling, high-resolution diffusion modeling can be both effective and straightforward.

Findings

Adjusted noise schedule improves high-res performance

Selective architecture scaling suffices for quality

Downsampling helps manage high-res feature maps

Abstract

Currently, applying diffusion models in pixel space of high resolution images is difficult. Instead, existing approaches focus on diffusion in lower dimensional spaces (latent diffusion), or have multiple super-resolution levels of generation referred to as cascades. The downside is that these approaches add additional complexity to the diffusion framework. This paper aims to improve denoising diffusion for high resolution images while keeping the model as simple as possible. The paper is centered around the research question: How can one train a standard denoising diffusion models on high resolution images, and still obtain performance comparable to these alternate approaches? The four main findings are: 1) the noise schedule should be adjusted for high resolution images, 2) It is sufficient to scale only a particular part of the architecture, 3) dropout should be added at specific locations in the architecture, and 4) downsampling is an effective strategy to avoid high resolution feature maps. Combining these simple yet effective techniques, we achieve state-of-the-art on image generation among diffusion models without sampling modifiers on ImageNet.