INVERSEBENCH
收藏资源简介:
INVERSEBENCH是由加州理工学院研究者创建的一个数据集,旨在评估基于插播扩散先验的逆问题算法。该数据集包含五个来自不同科学领域的逆问题,涵盖了从光学成像到流体动力学等多个领域,每个问题都呈现独特的结构性挑战。数据集通过定义不同类型的物理模型作为前向模型,展示了算法在不同噪声类型、线性与非线性问题上的性能。INVERSEBENCH为开源框架,提供了代码库、数据集和预训练模型,以促进相关领域的研究与发展。
INVERSEBENCH is a dataset developed by researchers at the California Institute of Technology, designed to evaluate inverse problem algorithms based on interpolated diffusion priors. This dataset includes five inverse problems spanning diverse scientific fields, ranging from optical imaging to fluid dynamics, with each problem presenting unique structural challenges. By defining different types of physical models as forward models, the dataset enables the assessment of algorithm performance across various noise types, linear and nonlinear inverse problems. INVERSEBENCH is an open-source framework that provides code repositories, datasets, and pretrained models to advance research and development in relevant domains.

- 1InverseBench: Benchmarking Plug-and-Play Diffusion Priors for Inverse Problems in Physical Sciences加州理工学院 · 2025年



