Single-Step Reconstruction-Free Anomaly Detection and Segmentation via Diffusion Models

1 st Mehrdad Moradi, 2 nd Marco Grasso, 3 rd Bianca Maria Colosimo, 4 th Kamran Paynabar · 2025

Read Single-Step Reconstruction-Free Anomaly Detection and Segmentation via Diffusion Models by 1 st Mehrdad Moradi, 2 nd Marco Grasso, 3 rd Bianca Maria Colosimo, 4 th Kamran Paynabar (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-2508-04818.txt.

Description

Moradi, Grasso, Colosimo, and Paynabar speed up the hunt for defects in images by skipping a slow step. Many systems spot flaws by rebuilding what a normal image should look like and comparing, but rebuilding takes many rounds and can drift to the wrong pattern. Their 2025 method, RADAR, reads the flaw map straight from a diffusion model in a single step.

Tests used a standard defect data set plus images of 3D-printed material. RADAR beat the best rival systems on every main score the authors report. Its top score rose by seven percent on the standard set and thirteen percent on the printed material set. The authors share their code, and note the preprint has not yet passed peer review.

Catalog record

CREATOR1 st Mehrdad Moradi, 2 nd Marco Grasso, 3 rd Bianca Maria Colosimo, 4 th Kamran Paynabar
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-04818

Browse the full library · About the Tower of Babel