Towards Training-free Anomaly Detection with Vision and Language Foundation Models | Read Paper on Bytez