Anomalous Samples for Few-Shot Anomaly Detection

2025-07-31

AI Summary

To address the scarcity of anomalous samples and the over-reliance of conventional few-shot anomaly detection methods on abundant normal data, this paper proposes a multi-score detection framework integrating zero-shot learning with memory augmentation. The method explicitly incorporates limited anomalous samples into training and introduces a data-augmentation-based validation mechanism to adaptively fuse three complementary scores: zero-shot anomaly score, memory bank matching score, and augmented classification score. This paradigm shift moves beyond normal-sample-dominated modeling, enabling robust representation learning for rare anomalies. Evaluated on mainstream industrial benchmarks including MVTec AD, the approach achieves significant improvements in detection accuracy and generalization under extreme few-shot settings (1–5 anomalous samples per class), with average AUC gains exceeding 12%.

Technology Category

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

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Several anomaly detection and classification methods rely on large amounts of non-anomalous or "normal" samples under the assump- tion that anomalous data is typically harder to acquire. This hypothesis becomes questionable in Few-Shot settings, where as little as one anno- tated sample can make a significant difference. In this paper, we tackle the question of utilizing anomalous samples in training a model for bi- nary anomaly classification. We propose a methodology that incorporates anomalous samples in a multi-score anomaly detection score leveraging recent Zero-Shot and memory-based techniques. We compare the utility of anomalous samples to that of regular samples and study the benefits and limitations of each. In addition, we propose an augmentation-based validation technique to optimize the aggregation of the different anomaly scores and demonstrate its effectiveness on popular industrial anomaly detection datasets.

Problem

• Utilizing anomalous samples for few-shot anomaly detection
• Comparing utility of anomalous vs normal samples in training
• Optimizing anomaly score aggregation via augmentation validation

Innovation

• Utilizes anomalous samples in training
• Multi-score detection with Zero-Shot techniques
• Augmentation-based validation for score optimization