To use the code in this repository first make a conda envoronment using environment.yml file.
Then prepare the data from AWE dataset (you should download it beforehand) with script data_prepare.py.
CNN training is implemented in script train.py, the best models are in folder best models.
You can test them using script test.py.
In notebook plots.ipynb you can see the selection of best models and empirical choosing of training parameters.
If you are interested in more information about training and testing, read the report ZrimsekUrsa3.pdf