A High-Quality Robust Diffusion Framework for Corrupted Dataset
This addresses the need for robust image generation in scenarios with corrupted data, representing an incremental advancement by adapting existing methods from GANs to diffusion models.
The authors tackled the problem of making diffusion models robust to outliers in training datasets, introducing the first robust-to-outlier diffusion framework that also improves performance on clean datasets.
Developing image-generative models, which are robust to outliers in the training process, has recently drawn attention from the research community. Due to the ease of integrating unbalanced optimal transport (UOT) into adversarial framework, existing works focus mainly on developing robust frameworks for generative adversarial model (GAN). Meanwhile, diffusion models have recently dominated GAN in various tasks and datasets. However, according to our knowledge, none of them are robust to corrupted datasets. Motivated by DDGAN, our work introduces the first robust-to-outlier diffusion. We suggest replacing the UOT-based generative model for GAN in DDGAN to learn the backward diffusion process. Additionally, we demonstrate that the Lipschitz property of divergence in our framework contributes to more stable training convergence. Remarkably, our method not only exhibits robustness to corrupted datasets but also achieves superior performance on clean datasets.