1 paper
Xurui Li, Enjie Xu, Chenzhou Li +4
Multi-class anomaly classification in industrial vision remains challenging due to noisy/incomplete anomaly representations and the unknown number of anomaly classes. To overcome t…