Unified Directly Denoising for Both Variance Preserving and Variance Exploding Diffusion Models

TL;DR

This paper introduces a unified framework for directly denoising both Variance Preserving and Variance Exploding diffusion models, achieving state-of-the-art image generation quality with fewer steps and theoretical guarantees.

Contribution

The paper proposes uDDDM, a unified model for VP and VE diffusion, with theoretical proofs and an adaptive loss function, improving efficiency and performance.

Findings

uDDDM achieves competitive FID scores on CIFAR-10.

One-step generation with FID of 2.63 (VE) and 2.53 (VP).

Extended sampling reduces FID to 1.71 (VE) and 1.65 (VP).

Abstract

Previous work has demonstrated that, in the Variance Preserving (VP) scenario, the nascent Directly Denoising Diffusion Models (DDDM) can generate high-quality images in one step while achieving even better performance in multistep sampling. However, the Pseudo-LPIPS loss used in DDDM leads to concerns about the bias in assessment. Here, we propose a unified DDDM (uDDDM) framework that generates images in one-step/multiple steps for both Variance Preserving (VP) and Variance Exploding (VE) cases. We provide theoretical proofs of the existence and uniqueness of the model's solution paths, as well as the non-intersecting property of the sampling paths. Additionally, we propose an adaptive Pseudo-Huber loss function to balance the convergence to the true solution and the stability of convergence process.Through a comprehensive evaluation, we demonstrate that uDDDMs achieve FID scores comparable to the best-performing methods available for CIFAR-10 in both VP and VE. Specifically, uDDDM achieves one-step generation on CIFAR10 with FID of 2.63 and 2.53 for VE and VP respectively. By extending the sampling to 1000 steps, we further reduce FID score to 1.71 and 1.65 for VE and VP respectively, setting state-of-the-art performance in both cases.