Anomaly Detection and Localization in Optical Networks Using Vision Transformer and SOP Monitoring

We introduce an innovative vision transformer approach to identify and precisely locate high-risk events, including fiber cut precursors, in state-of-polarization derived spectrograms. Our method achieves impressive 97% diagnostic accuracy and precise temporal localization (6-ms- RMSE).