Patch distribution modeling framework adaptive cosine estimator (PaDiM-ACE) for anomaly detection and localization in synthetic aperture radar imagery

Angelina Ibarra, Joshua Peeples · 2025

Read Patch distribution modeling framework adaptive cosine estimator (PaDiM-ACE) for anomaly detection and localization in synthetic aperture radar imagery by Angelina Ibarra, Joshua Peeples (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-08049.txt.

Description

Angelina Ibarra and Joshua Peeples adapt an anomaly hunting method for radar images that see through clouds and darkness. Synthetic aperture radar pictures look grainy and strange, so spotting the odd patch takes care. The base method, PaDiM, scores each patch by how far it sits from learned normal patterns, but that score has no ceiling.

Their variant swaps in the adaptive cosine estimator, which judges by angle instead of distance. Scores now live inside fixed bounds, making thresholds easier to set and compare. The pair test the change across several radar data sets, scoring hits at both whole-image and pixel level. Code is public.

Catalog record

CREATORAngelina Ibarra, Joshua Peeples
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE18 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-08049

Browse the full library · About the Tower of Babel