Reproduce “SGDB: Semantic-Geometric Decoupling and Evidence Binding for Logical Anomaly Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Chenrui Ji, Jiafan Lu, Zilu Luo, Feng Wan, Qiuhui Chen, Weimin Zhong
SGDB detects logical anomalies by separating object recognition from geometric measurement, then binding measurements to identities. Its support-certified length reader accepts mask-derived axes only when line quality and complete-component ownership checks pass; unmatched identities remain counted, with geometry unavailable. It reports task-macro AUROC/F1 of 98.19%/94.52% on MVTec LOCO and 93.56%/92.76% across 50 VID-AD tasks, in a supervised, configured inspection setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.