Abstract: Applying machine vision to facilitate railway anomaly detections faces a grand challenge in that anomalous samples for model training are insufficient due to their infrequent occurrence and ...