Physical World & Real-World ImpactUniversity of Padua (Barusco, Dalle Pezze, Susto et al.)
University of Padua compares anomaly detection on integrated circuit SEM images
Compared anomaly detection on integrated circuit SEM images — 25,160 normal and 116 anomalous images. By the authors' measurement, CFA reached an image-AUROC of 99.01% and image-F1 of 87.96%, though the best pixel-F1 was 77.30%
Why it matters
99% image-level AUROC with zero defect labels. A public comparative experiment showing that, for OSAT, where defect classes change often, a design that skips the defect-collection stage entirely holds up in the SEM domain
Source — Primary sources first; secondary reports are labelled “Press”
Paperarxiv.orghttps://arxiv.org/html/2505.07576v1Added in Week 35, 2026 · See that week