arXiv:2605.27748v1 Announce Type: cross 
Abstract: Industrial visual anomaly detection is usually one-class: normal images are abundant, while defects are rare, heterogeneous, and often unavailable during system design. PatchCore-style retrieval suits this setting because it scores test images from a memory bank of normal patch features, but the standard Euclidean geometry ignores feature correlations and its offline construction materialises the full patch pool before subsampling.
  We introduce Mahalanobis PatchCore, a covariance-aware, streaming-compatible extension of PatchCore. Its artificial intelligence contribution is a retrieval detector that estimates a regularised covariance model in reduced feature space and whitens embeddings, so Euclidean nearest-neighbour search after transformation implements Mahalanobis retrieval. A bounded-memory, re-iterable training pipeline builds the memory bank without storing all normal patches at once, using incremental dimensionality reduction, online covariance estimation, and streaming aggregation.
  The engineering application is automated industrial inspection, where visual anomaly detection must remain accurate under practical memory limits. We evaluate the method on a public 15-category industrial anomaly-detection benchmark and three industrial datasets covering blow-fill-seal strip-ampoule meniscus inspection, amber-glass-ampoule bottom inspection, and lyophilised-cake vial inspection. Mahalanobis PatchCore preserves most offline PatchCore image-level performance on the public benchmark while reducing peak memory from 5.41 to 2.78 GB, and improves the selected industrial mean image area under the receiver operating characteristic curve from 0.981 to 0.986.

يمثل إدخال Mahalanobis PatchCore تقدمًا كبيرًا في الكشف عن الشذوذ البصري الصناعي، حيث يعالج القيود التي تفرضها طرق الكشف التقليدية ذات الفئة الواحدة. يتضمن هذا الإطار الجديد الوعي بالتغاير وملاءمة البث، مما يسمح باسترجاع أكثر فعالية لميزات التصحيح العادية من بنك الذاكرة دون الحاجة إلى تخزين بيانات خارجية واسعة.

La introducción de Mahalanobis PatchCore marca un avance significativo en la detección visual de anomalías industriales, abordando las limitaciones de los métodos tradicionales de detección de una sola clase. Este nuevo marco incorpora la conciencia de la covarianza y la compatibilidad con el streaming, lo que permite una recuperación más efectiva de características de parches normales de un banco de memoria sin necesidad de un almacenamiento de datos fuera de línea extenso.

L'introduction de Mahalanobis PatchCore représente une avancée significative dans la détection visuelle des anomalies industrielles, répondant aux limites des méthodes de détection à une seule classe. Ce nouveau cadre intègre la prise en compte de la covariance et la compatibilité avec le streaming, permettant une récupération plus efficace des caractéristiques de patch normales à partir d'une banque de mémoire sans nécessiter un stockage de données hors ligne étendu.

The introduction of Mahalanobis PatchCore marks a significant advancement in industrial visual anomaly detection, addressing the limitations of traditional one-class detection methods. This new framework incorporates covariance awareness and streaming compatibility, enabling more effective retrieval of normal patch features from a memory bank without the need for extensive offline data storage.