Background-Aware Defect Generation for Robust Industrial Anomaly Detection

2024-11-25

AI Summary

Industrial anomaly detection suffers from severe scarcity of labeled anomalous samples, particularly impeding semantic modeling of interactions between defects and their background contexts—leading generative methods to exhibit insufficient realism and contextual consistency in logically complex anomalies. This paper proposes a background-aware defect generation framework featuring a novel background-defect disentangled denoising mechanism, theoretically guaranteeing structural fidelity of the background while enabling controllable defect synthesis. By integrating DDIM inversion with inverse latent-space editing, the framework supports background-guided conditional generation. Evaluated on MVTec AD and MVTec LOCO, our method significantly improves both the realism and logical consistency of synthesized defects. Consequently, downstream anomaly detection achieves higher detection rates and localization accuracy—especially pronounced in challenging logical anomaly scenarios.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

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 learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Detecting anomalies in industrial settings is challenging due to the scarcity of labeled anomalous data. Generative models can mitigate this issue by synthesizing realistic defect samples, but existing approaches often fail to model the crucial interplay between defects and their background. This oversight leads to unrealistic anomalies, especially in scenarios where contextual consistency is essential (i.e., logical anomaly). To address this, we propose a novel background-aware defect generation framework, where the background influences defect denoising without affecting the background itself by ensuring realistic synthesis while preserving structural integrity. Our method leverages a disentanglement loss to separate the background' s denoising process from the defect, enabling controlled defect synthesis through DDIM Inversion. We theoretically demonstrate that our approach maintains background fidelity while generating contextually accurate defects. Extensive experiments on MVTec AD and MVTec Loco benchmarks validate our mehtod's superiority over existing techniques in both defect generation quality and anomaly detection performance.

Problem

• Addresses scarcity of labeled anomalous data in industrial anomaly detection.
• Improves defect synthesis by modeling defect-background interplay for realistic anomalies.
• Ensures contextual consistency and structural integrity in generated defects.

Innovation

• Background-aware defect generation framework
• Disentanglement loss separates background and defect
• DDIM Inversion for controlled defect synthesis