Self-Supervised Anomaly Detection from Anomalous Training Data via Iterative Latent Token Masking | IEEE Conference Publication | IEEE Xplore