🤖 AI Summary
This work addresses the challenge of industrial anomaly detection under contaminated training data—where normal samples are inadvertently mixed with anomalies—a scenario that undermines the performance of existing methods relying on either purely clean or entirely unlabeled data. To tackle this, we propose a robust anomaly detection framework that leverages a small number of annotated anomalous samples under contamination. Our approach introduces adaptive bias learning and a tripartite anomaly scoring mechanism that integrates statistical deviation, entropy-based uncertainty, and segmentation-space anomaly cues to enable accurate detection and interpretable localization. Coupled with gradient-guided localization and adaptive instance weighting, the method achieves state-of-the-art performance on the MVTec and VisA benchmarks, demonstrating consistently high detection accuracy, precise localization, and strong robustness across varying contamination levels.
📝 Abstract
Visual anomaly detection in real-world industrial settings faces two major limitations. First, most existing methods are trained on purely normal data or on unlabeled datasets assumed to be predominantly normal, presuming the absence of contamination, an assumption that is rarely satisfied in practice. Second, they assume no access to labeled anomaly samples, limiting the model from learning discriminative characteristics of true anomalies. Therefore, these approaches often struggle to distinguish anomalies from normal instances, resulting in reduced detection and weak localization performance. In real-world applications, where training data are frequently contaminated with anomalies, such methods fail to deliver reliable performance. In this work, we propose a robust anomaly detection framework that integrates limited anomaly supervision into the adaptive deviation learning paradigm. We introduce a composite anomaly score that combines three complementary components: a deviation score capturing statistical irregularity, an entropy-based uncertainty score reflecting predictive inconsistency, and a segmentation-based score highlighting spatial abnormality. This unified scoring mechanism enables accurate detection and supports gradient-based localization, providing intuitive and explainable visual evidence of anomalous regions. Following the few-anomaly paradigm, we incorporate a small set of labeled anomalies during training while simultaneously mitigating the influence of contaminated samples through adaptive instance weighting. Extensive experiments on the MVTec and VisA benchmarks demonstrate that our framework outperforms state-of-the-art baselines and achieves strong detection and localization performance, interpretability, and robustness under various levels of data contamination.