arxivcs.CV2026-06-28
Robust Zero-shot Anomaly Detection under Limited Auxiliary Anomaly Priors
Guanyu Lu, Fang Zhou, Cheqing Jin
Zero-shot anomaly detection aims to identify defects in arbitrary novel domains; however, existing models assume that the auxiliary data contains a rich diversity of anomalies, neglecting the far more complex and unpredictable variations in real-world target domains. This study i…