Python Code to Train a Neural Network for the Identification of EMIC Wave Events in Spectrograms

• Main Author: Nyssa Capman
• Format: Other/Unknown Material
• Language: English
• Published: Zenodo 2023
• Subjects: Space physics; Electromagnetic ion cyclotron (EMIC) waves; Spectrograms; Antarc*; Antarctica
• Online Access: https://doi.org/10.5281/zenodo.8280090

Python code used to train a convolutional neural network (CNN) for the identification of electromagnetic ion cyclotron (EMIC) wave events in spectrograms. Three versions of the code are provided: one to train and test a model (CNN_Training_zenodo.py), one to test a pre-trained model on new, labeled test data (CNN_Testing_zenodo.py), and one to test a pre-trained model on new, unlabeled test data (CNN_Classify_zenodo.py). Additionally, pre-labeled spectrograms from the Halley, Antarctica ground magnetometer station between November 2006 and December 2009 are provided, as well as lookup tables to relate pixel locations to time and frequency information in the provided spectrograms.