The plotting notebook contains all plots used in the thesis. I will try to add the models referenced in this notebook as well, so it can be rerun by anyone. Since the models are rather large, only the validation scores will be added.
The BAtorch notebook is the main work of the thesis. It contains the implementation of the CJRCWrapper, SELayer, MBConv, LegNet and EfficientNetV2. The trainer, as well as LegNet is largely based on their code, only small adjustments were made. The notebook also contains the evaluation with the left-out test data, as well as the cross validation code.
All supporting functionality have been moved to the helper_functions.py.
The train_sequences, as well as the "filtered_test_data_with_MAUDE_expressions" test data can be found at https://zenodo.org/records/7395397.