Reproduce “Beyond Geometry: Benchmarking and Consistency Reasoning for 3D 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
Zhiqiang Qin, He Xie, Junfei Yi, Yang Yang, Hao Wang, Yunkang Cao, Hui Zhang, Yaonan Wang
ILGAD introduces a benchmark for detecting structural and assembly-rule violations in industrial 3D point clouds, alongside a method that checks local geometry, expected structure coverage, and spatial relationships. The method reaches 91.90% object-wise and 96.28% point-wise AUROC on ILGAD, outperforming the strongest mean baseline by 15.99 and 6.79 percentage points, respectively.
Key insights
- Logical anomalies may be evident only by comparing an object with its expected structure; detecting missing parts requires reasoning about reference geometry that is absent from the observation.
- The three consistency signals address complementary evidence: geometry detects visible defects, coverage identifies unsupported or missing structures, and spatial relations flag abnormal arrangements.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.