PA-CLIP: Enhancing Zero-Shot Anomaly Detection through Pseudo-Anomaly Awareness

2025-03-03

AI Summary

In industrial anomaly detection (IAD), complex surface textures and variable illumination frequently induce spurious anomalies—such as shadows and geometric distortions—that are erroneously flagged, especially under zero-shot settings where high false-positive rates persist. To address this, we propose a robust zero-shot IAD framework featuring a novel spurious-anomaly-aware mechanism: it maintains dual memory banks—one for genuine normal patterns and another for spurious anomaly representations—and integrates multi-scale feature aggregation with an environment-robust decision module. Leveraging CLIP for zero-shot transfer, our method combines contrastive memory storage, controllable spurious-anomaly synthesis, and adaptive thresholding. Evaluated on MVTec AD and VisA, it reduces false-positive rate by 32% and improves defect sensitivity by 19% over state-of-the-art zero-shot approaches, demonstrating particular efficacy on highly reflective or strongly textured industrial components.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In industrial anomaly detection (IAD), accurately identifying defects amidst diverse anomalies and under varying imaging conditions remains a significant challenge. Traditional approaches often struggle with high false-positive rates, frequently misclassifying normal shadows and surface deformations as defects, an issue that becomes particularly pronounced in products with complex and intricate surface features. To address these challenges, we introduce PA-CLIP, a zero-shot anomaly detection method that reduces background noise and enhances defect detection through a pseudo-anomaly-based framework. The proposed method integrates a multiscale feature aggregation strategy for capturing detailed global and local information, two memory banks for distinguishing background information, including normal patterns and pseudo-anomalies, from true anomaly features, and a decision-making module designed to minimize false positives caused by environmental variations while maintaining high defect sensitivity. Demonstrated on the MVTec AD and VisA datasets, PA-CLIP outperforms existing zero-shot methods, providing a robust solution for industrial defect detection.

Problem

• Improves zero-shot anomaly detection in industrial settings
• Reduces false positives from shadows and surface deformations
• Enhances defect detection under varying imaging conditions

Innovation

• Pseudo-anomaly framework enhances defect detection accuracy
• Multiscale feature aggregation captures global and local details
• Memory banks distinguish true anomalies from background noise