This is the preprocessing part of our solution. The scripts are to be run in order. The shell scripts should be run in the working directory (under which the dataset is downloaded), while the python scripts 2, 3, and 4 accept a parameter pointing to this directory.
The scripts 5 and 6 need to be run for each "split folder" separately, with their one command line argument pointed the directory. The current setup creates two split folders ("full" and "0").