KLIP: localized distribution shift detection via KL-divergence with diffusion priors in Inverse Problems
2026-05-29 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors studied how to find images that don’t fit the usual patterns (called out-of-distribution or OOD) using diffusion models, which are tools that generate or understand images. They created a new way to measure difference using something called the Kullback-Leibler divergence that doesn’t need extra example data or prior knowledge about unusual images. Their method can spot whole unusual images as well as specific odd parts inside images, even when the differences are small but important, like spotting tumors in liver scans. They tested it on different models, data, and imaging problems, showing it works broadly.
Diffusion modelsOut-of-distribution detectionKullback-Leibler divergenceComputational imagingInverse problemsPosterior distributionPrior distributionCT scansDistribution shiftsImage localization
Authors
Alireza Kheirandish, Jihoon Hong, Sara Fridovich-Keil
Abstract
Diffusion models have shown promising performance as data-driven priors for computational imaging, as well as some capacity to detect out-of-distribution (OOD) images. However, existing approaches to OOD detection often require some knowledge of the shifted distribution, fail to detect subtle or localized distribution shifts, and operate on full images, rather than the indirect measurements available in inverse problems. We propose an OOD detection metric based on the Kullback-Leibler divergence between the diffusion prior and the posterior distribution, that (i) does not require any calibration data or knowledge of the shifted distribution, and (ii) can detect whole images as OOD as well as localize OOD patches within an image. Experimentally, we show that this metric can detect subtle yet semantically meaningful distribution shifts, such as the shift from healthy liver CT scans to those with tumors, and generalizes across different types of diffusion models, datasets, and inverse problems. Our code can be found at https://github.com/voilalab/KLIP.