Unsupervised Anomaly Detection Using Diffusion Trend Analysis

2024-07-12

arXiv.org

AI Summary

To address two key challenges in applying denoising diffusion models to industrial unsupervised anomaly detection—difficulty in adapting noise parameters and false positives caused by reconstruction instability—this paper proposes a diffusion-trend-aware anomaly detection method. Our approach introduces, for the first time, multi-scale reconstruction trajectory modeling over the diffusion process, explicitly decoupling the progressive degradation of anomalous regions from the robust preservation of normal structures—thereby eliminating reliance on fixed noise schedules or hand-crafted residual thresholds. We further design an unsupervised pixel-level anomaly scoring function that jointly ensures localization robustness and image-level discriminative reliability. Evaluated on standard industrial benchmark datasets, our method achieves significant improvements in AUROC and PRO scores, while maintaining high accuracy and low computational overhead—enabling real-time deployment on production lines.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Conventional anomaly detection techniques based on reconstruction via denoising diffusion model are widely used due to their ability to identify anomaly locations and shapes with high performance. However, there is a limitation in determining appropriate noise parameters that can degrade anomalies while preserving normal characteristics. Also, due to the volatility of the diffusion model, normal regions can fluctuate considerably during reconstruction, resulting in false detection. In this paper, we propose a method to detect anomalies by analysis of reconstruction trend depending on the degree of degradation, effectively solving the both problems of existing methods. The proposed method is validated on an open dataset for industrial anomaly detection, improving the performance of existing methods on a number of evaluation criteria. With the ease of combination with existing anomaly detection methods, it provides a tradeoff between computational cost and performance, allowing it high application potential in manufacturing industry.

Problem

• Determining optimal noise parameters for anomaly detection
• Reducing false detection from normal region fluctuations
• Improving display inspection via reconstruction trend analysis

Innovation

• Uses diffusion trend analysis for anomaly detection
• Analyzes reconstruction trend by degradation degree
• Solves noise parameter and fluctuation issues