GSURE-Based Diffusion Model Training with Corrupted Data

arXiv:2305.13128v2 Announce Type: replace-cross
Abstract: Diffusion models have demonstrated impressive results in both data generation and downstream tasks such as inverse problems, text-based editing, classification, and more. However, training such models usually requires large amounts of clean signals which are often difficult or impossible to obtain. In this work, we propose a novel training technique for generative diffusion models based only on corrupted data. We introduce a loss function based on the Generalized Stein's Unbiased Risk Estimator (GSURE), and prove …

abstract arxiv classification corrupted data cs.cv cs.lg data data generation diffusion diffusion model diffusion models editing eess.iv however novel replace results tasks text training type work