Memory-Distilled Selection for Noise-Robust Anomaly Detection

Anomaly detection (AD) under data contamination is critical for deploying unsupervised defect detection in industrial environments, where curating perfectly clean training sets is impractical. However, existing methods are sensitive to contamination, suffering significant performance degradation as the noise ratio increases. In this paper, we propose Memory-Distilled Selection (MeDS), a training algorithm based on data selection. MeDS constructs an ensemble of partial memories via random subsampling, where the resulting sparsity acts as a low-pass filter that captures nominal patterns across a wide range of noise ratios, enabling coarse-level identification of contaminated samples. The aggregated distances to the bootstrapped memories are then distilled into a reconstruction score network, which is subsequently fine-tuned on clean data filtered using scores from the distilled model, enabling fine-grained localization of anomalies. MeDS is robust across a wide range of noise ratios without requiring noise-ratio-specific hyperparameter tuning, achieving 99.16\% image-level AUROC on MVTecAD at a 40\% noise ratio, and attaining state-of-the-art performance on both VisA and Real-IAD under noisy settings. We thoroughly verify the efficacy of MeDS on industrial AD benchmarks under noisy data scenarios, accompanied by in-depth empirical analyses.

Detecting defective products in manufacturing requires systems that learn what "normal" looks like from training data. In practice, this training data is often imperfect, as some defective items are accidentally included as normal examples. Existing detection methods struggle as the fraction of such mislabeled data grows. We propose a training method that builds many small, partial models of normality through random sampling, which naturally filters out the influence of mislabeled examples. These partial models are then combined into a single scoring system, which is used to identify and remove contaminated samples, allowing the system to precisely locate defects. Our method works reliably across a wide range of contamination levels without needing to know how much of the data is mislabeled, and achieves state-of-the-art results on multiple industrial inspection benchmarks.