This code consists in three main python files:
load_data.py
that load the training tracks and perform some data augmentation;build_autoencoder.py
that specifies the autoencoder architecture, the way in which the autoencoder is used to predict new points and train the model;GT-tracks_filtering.py
that test the model on ground truth data and returns the filtered track and the RMSD between the filtered and the true trajectory.
The testing can be performed over a specific track passing the name as parameter using the sintax python GT-tracks_filtering --track "track_name"
. A shell file is added to automatize the work over the whole testing tracks returning also a file containing the RMSE computed over each track called metrics.txt
.