Towards Generalizable 3D Anomaly Detection via Relational Inconsistency Modeling

2026-09-28

AI Summary

This study addresses the blurred boundaries and high false positive rates in existing 3D anomaly detection caused by the absence of explicit defect modeling. To overcome these limitations, this work proposes a relational inconsistency modeling framework that departs from the conventional normality distribution paradigm by defining defects as violations of local geometric structural relationships, extracting category-agnostic defect features via pseudo-anomaly learning. Methodologically, an edge-aware graph refinement (EGR) module is designed to encode geometric relationships, coupled with cluster deviation modeling (CDM) to precisely localize structurally incompatible regions. Experimental results demonstrate that the proposed approach significantly outperforms state-of-the-art methods on benchmarks such as Anomaly-ShapeNet, achieving superior performance in both intra-domain and cross-domain detection scenarios.

Abstract

3D anomaly detection (3DAD) aims to identify defective regions in point cloud data, serving as a critical component in industrial inspection systems. Existing methods are normality-centered -- learning the distribution of normal samples and treating deviations as anomalies -- without explicitly modeling what constitutes a defect. This leads to ambiguous decision boundaries with increased false positives and negatives, particularly in unified and cross-domain settings where diverse normal distributions further blur the boundaries. We propose a relational inconsistency modeling framework that characterizes defects as violations of geometric consistency among neighboring structures. Our approach learns category-agnostic defect cues through pseudo-anomalies designed as controlled relational violations, instantiated by two key modules: Edge-aware Graph Refinement (EGR) for encoding geometric relationships among local regions, and Cluster-Deviation Modeling (CDM) for identifying regions that are relationally incompatible within their structural peer group. Extensive experiments on Anomaly-ShapeNet and Real3D-AD demonstrate consistent improvements over prior state-of-the-art methods in both in-domain and cross-domain settings, validating the effectiveness of learning an explicit, relation-based defect criterion for 3D anomaly detection. Project page: https://visualsciencelab-khu.github.io/GRIM_project/.

Problem

• 3D anomaly detection
• cross-domain generalization
• point cloud
• decision boundary ambiguity
• defect modeling

Innovation

• 3D Anomaly Detection
• Relational Inconsistency Modeling
• Pseudo-anomalies
• Edge-aware Graph Refinement
• Cluster-Deviation Modeling