mixed attention auto encoder for multi-class industrial anomaly detection

Jiangqi Liu, Feng Wang

Most existing methods for unsupervised industrial anomaly detection train a separate model for each object category. This kind of approach can easily capture the category-specific…