Center-aware Residual Anomaly Synthesis for Multi-class Industrial Anomaly Detection

2025-05-23

AI Summary

Industrial multi-class unified anomaly detection faces two key challenges: inter-class interference causing missed detections, and feature-space overlap between normal and anomalous samples within the same class leading to over-detection. To address these, we propose a center-aware residual learning mechanism that aligns multi-class features toward a shared latent center; a distance-guided anomaly synthesis strategy that adaptively modulates noise variance to suppress intra-class ambiguity; and multi-class feature decoupling jointly optimized with variance-adaptive noise injection calibrated to the normal distribution. Our method achieves significant improvements in both detection accuracy and inference speed across multiple industrial benchmarks and real-world production lines. The source code and a newly curated multi-class industrial anomaly detection benchmark dataset are publicly released.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Multi-modal Vision

Application Category

• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Anomaly detection plays a vital role in the inspection of industrial images. Most existing methods require separate models for each category, resulting in multiplied deployment costs. This highlights the challenge of developing a unified model for multi-class anomaly detection. However, the significant increase in inter-class interference leads to severe missed detections. Furthermore, the intra-class overlap between normal and abnormal samples, particularly in synthesis-based methods, cannot be ignored and may lead to over-detection. To tackle these issues, we propose a novel Center-aware Residual Anomaly Synthesis (CRAS) method for multi-class anomaly detection. CRAS leverages center-aware residual learning to couple samples from different categories into a unified center, mitigating the effects of inter-class interference. To further reduce intra-class overlap, CRAS introduces distance-guided anomaly synthesis that adaptively adjusts noise variance based on normal data distribution. Experimental results on diverse datasets and real-world industrial applications demonstrate the superior detection accuracy and competitive inference speed of CRAS. The source code and the newly constructed dataset are publicly available at https://github.com/cqylunlun/CRAS.

Problem

• Unified model needed for multi-class anomaly detection to reduce deployment costs
• Inter-class interference causes severe missed detections in anomaly detection
• Intra-class overlap between normal and abnormal samples leads to over-detection

Innovation

• Center-aware residual learning for unified multi-class detection
• Distance-guided anomaly synthesis to reduce overlap
• Adaptive noise variance based on data distribution