Code extract for submission #1733 On the Effects of Data Distortion on Model Analysis and Training

The main experiments can be reproduced by running the DI.py and iOcclusion.py python scripts.
DI.py computes the Data Interference (DI) index and takes as arguments the model path and the local path to the data set.
iOcclusion.py outputs the iOcclusion measure and takes as arguments an occlusion percentage (float), model and data set paths.
A complete version of the code that includes all experiments will be released upon acceptance.
