InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zheng, Hongkai, Chu, Wenda, Zhang, Bingliang, Wu, Zihui, Wang, Austin, Feng, Berthy T., Zou, Caifeng, Sun, Yu, Kovachki, Nikola, Ross, Zachary E., Bouman, Katherine L., Yue, Yisong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909815791419392
author Zheng, Hongkai
Chu, Wenda
Zhang, Bingliang
Wu, Zihui
Wang, Austin
Feng, Berthy T.
Zou, Caifeng
Sun, Yu
Kovachki, Nikola
Ross, Zachary E.
Bouman, Katherine L.
Yue, Yisong
author_facet Zheng, Hongkai
Chu, Wenda
Zhang, Bingliang
Wu, Zihui
Wang, Austin
Feng, Berthy T.
Zou, Caifeng
Sun, Yu
Kovachki, Nikola
Ross, Zachary E.
Bouman, Katherine L.
Yue, Yisong
contents Plug-and-play diffusion priors (PnPDP) have emerged as a promising research direction for solving inverse problems. However, current studies primarily focus on natural image restoration, leaving the performance of these algorithms in scientific inverse problems largely unexplored. To address this gap, we introduce \textsc{InverseBench}, a framework that evaluates diffusion models across five distinct scientific inverse problems. These problems present unique structural challenges that differ from existing benchmarks, arising from critical scientific applications such as optical tomography, medical imaging, black hole imaging, seismology, and fluid dynamics. With \textsc{InverseBench}, we benchmark 14 inverse problem algorithms that use plug-and-play diffusion priors against strong, domain-specific baselines, offering valuable new insights into the strengths and weaknesses of existing algorithms. To facilitate further research and development, we open-source the codebase, along with datasets and pre-trained models, at https://devzhk.github.io/InverseBench/.
format Preprint
id arxiv_https___arxiv_org_abs_2503_11043
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences
Zheng, Hongkai
Chu, Wenda
Zhang, Bingliang
Wu, Zihui
Wang, Austin
Feng, Berthy T.
Zou, Caifeng
Sun, Yu
Kovachki, Nikola
Ross, Zachary E.
Bouman, Katherine L.
Yue, Yisong
Machine Learning
Plug-and-play diffusion priors (PnPDP) have emerged as a promising research direction for solving inverse problems. However, current studies primarily focus on natural image restoration, leaving the performance of these algorithms in scientific inverse problems largely unexplored. To address this gap, we introduce \textsc{InverseBench}, a framework that evaluates diffusion models across five distinct scientific inverse problems. These problems present unique structural challenges that differ from existing benchmarks, arising from critical scientific applications such as optical tomography, medical imaging, black hole imaging, seismology, and fluid dynamics. With \textsc{InverseBench}, we benchmark 14 inverse problem algorithms that use plug-and-play diffusion priors against strong, domain-specific baselines, offering valuable new insights into the strengths and weaknesses of existing algorithms. To facilitate further research and development, we open-source the codebase, along with datasets and pre-trained models, at https://devzhk.github.io/InverseBench/.
title InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences
topic Machine Learning
url https://arxiv.org/abs/2503.11043