Core Concepts
Proposing DMAD framework for real-world anomaly detection, combining unsupervised and semi-supervised scenarios.
Abstract
Training a unified model is efficient for industrial anomaly detection.
DMAD uses dual memory banks to handle normal and abnormal patterns.
Evaluation on MVTec-AD and VisA datasets shows superior performance.
Stats
このフレームワークは、MVTec-ADおよびVisAデータセットで現在の最先端の手法を上回る結果を示しています。