Leveraging Learning Bias for Noisy Anomaly Detection

TL;DR

This paper introduces a two-stage, model-agnostic framework that exploits inherent learning biases to improve unsupervised image anomaly detection in noisy training data, achieving superior performance and robustness.

Contribution

The proposed method systematically leverages learning bias to filter contaminated training data, enhancing anomaly detection without requiring clean datasets.

Findings

Outperforms existing methods on Real-IAD benchmark

Demonstrates robustness to training data contamination

Compatible with various unsupervised backbones

Abstract

This paper addresses the challenge of fully unsupervised image anomaly detection (FUIAD), where training data may contain unlabeled anomalies. Conventional methods assume anomaly-free training data, but real-world contamination leads models to absorb anomalies as normal, degrading detection performance. To mitigate this, we propose a two-stage framework that systematically exploits inherent learning bias in models. The learning bias stems from: (1) the statistical dominance of normal samples, driving models to prioritize learning stable normal patterns over sparse anomalies, and (2) feature-space divergence, where normal data exhibit high intra-class consistency while anomalies display high diversity, leading to unstable model responses. Leveraging the learning bias, stage 1 partitions the training set into subsets, trains sub-models, and aggregates cross-model anomaly scores to filter a purified dataset. Stage 2 trains the final detector on this dataset. Experiments on the Real-IAD benchmark demonstrate superior anomaly detection and localization performance under different noise conditions. Ablation studies further validate the framework's contamination resilience, emphasizing the critical role of learning bias exploitation. The model-agnostic design ensures compatibility with diverse unsupervised backbones, offering a practical solution for real-world scenarios with imperfect training data. Code is available at https://github.com/hustzhangyuxin/LLBNAD.