DMAD: Dual Memory Bank for Real-World Anomaly Detection

TL;DR

DMAD introduces a dual memory bank framework that enhances anomaly detection by effectively leveraging both normal and abnormal data in a unified model, improving performance on real-world datasets.

Contribution

The paper proposes a novel dual memory bank approach for unified anomaly detection, integrating normal and abnormal knowledge for better representation learning.

Findings

DMAD outperforms state-of-the-art methods on MVTec-AD and VisA datasets.

The dual memory bank effectively captures normal and abnormal features.

The framework handles both unsupervised and semi-supervised scenarios.

Abstract

Training a unified model is considered to be more suitable for practical industrial anomaly detection scenarios due to its generalization ability and storage efficiency. However, this multi-class setting, which exclusively uses normal data, overlooks the few but important accessible annotated anomalies in the real world. To address the challenge of real-world anomaly detection, we propose a new framework named Dual Memory bank enhanced representation learning for Anomaly Detection (DMAD). This framework handles both unsupervised and semi-supervised scenarios in a unified (multi-class) setting. DMAD employs a dual memory bank to calculate feature distance and feature attention between normal and abnormal patterns, thereby encapsulating knowledge about normal and abnormal instances. This knowledge is then used to construct an enhanced representation for anomaly score learning. We evaluated DMAD on the MVTec-AD and VisA datasets. The results show that DMAD surpasses current state-of-the-art methods, highlighting DMAD's capability in handling the complexities of real-world anomaly detection scenarios.