This is the initial commit. Some of the code had some last minute changes to improve readability, and has not been tested afterwards. This might result in a minor error or two. A lot of optimization and code reduction still needs to be made.
The data files goes in the data folder. Look into data/index.json to understand
the current structuring or change it if needed. The datasets should only contain
files with accelerometer data.
The scripts to train the models are main/timedata_training.py and main/spectrum_training.py.
They will save the trained models to a main/saved_models/ folder. The training logs
will be saved in a main/logs/ folder. At the moment, these folder
must be manually created
The scripts to test the models are main/timedata_prediction.py and main/spectrum_prediction.py.
They will load the trained models from the main/saved_models/ folder.
They will also save the predictions, figures and statistics to the main/validation folder,
which must be manually created.