Saved in:
Bibliographic Details
Main Authors: Kou, Siqi, Gan, Lei, Wang, Dequan, Li, Chongxuan, Deng, Zhijie
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2310.11142
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910351997534208
author Kou, Siqi
Gan, Lei
Wang, Dequan
Li, Chongxuan
Deng, Zhijie
author_facet Kou, Siqi
Gan, Lei
Wang, Dequan
Li, Chongxuan
Deng, Zhijie
contents Diffusion models have impressive image generation capability, but low-quality generations still exist, and their identification remains challenging due to the lack of a proper sample-wise metric. To address this, we propose BayesDiff, a pixel-wise uncertainty estimator for generations from diffusion models based on Bayesian inference. In particular, we derive a novel uncertainty iteration principle to characterize the uncertainty dynamics in diffusion, and leverage the last-layer Laplace approximation for efficient Bayesian inference. The estimated pixel-wise uncertainty can not only be aggregated into a sample-wise metric to filter out low-fidelity images but also aids in augmenting successful generations and rectifying artifacts in failed generations in text-to-image tasks. Extensive experiments demonstrate the efficacy of BayesDiff and its promise for practical applications.
format Preprint
id arxiv_https___arxiv_org_abs_2310_11142
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle BayesDiff: Estimating Pixel-wise Uncertainty in Diffusion via Bayesian Inference
Kou, Siqi
Gan, Lei
Wang, Dequan
Li, Chongxuan
Deng, Zhijie
Computer Vision and Pattern Recognition
Machine Learning
Diffusion models have impressive image generation capability, but low-quality generations still exist, and their identification remains challenging due to the lack of a proper sample-wise metric. To address this, we propose BayesDiff, a pixel-wise uncertainty estimator for generations from diffusion models based on Bayesian inference. In particular, we derive a novel uncertainty iteration principle to characterize the uncertainty dynamics in diffusion, and leverage the last-layer Laplace approximation for efficient Bayesian inference. The estimated pixel-wise uncertainty can not only be aggregated into a sample-wise metric to filter out low-fidelity images but also aids in augmenting successful generations and rectifying artifacts in failed generations in text-to-image tasks. Extensive experiments demonstrate the efficacy of BayesDiff and its promise for practical applications.
title BayesDiff: Estimating Pixel-wise Uncertainty in Diffusion via Bayesian Inference
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2310.11142