The paper introduces BayesDiff, a framework for estimating pixel-wise uncertainty in diffusion models using Bayesian inference. It addresses challenges in image quality identification and generation enhancement. The framework is detailed with novel uncertainty iteration principles and efficient Bayesian inference strategies. Extensive experiments demonstrate the efficacy of BayesDiff in practical applications.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Siqi Kou,Lei... at arxiv.org 03-05-2024
https://arxiv.org/pdf/2310.11142.pdfDeeper Inquiries