The purpose of the Thesis was to try to understand the relevant features a Convolutional Neural Network (CNN) needs to learn to classify breast cancer types. Thus, did I first segment the images, classifying the regions in 4 semantic types. The classifier was then supposed to learn from the four colours, a much harder task considering that the information in the segemented images was much less than in the original images. The Segmentation folder contains the part for Segmentation, while the Classification folder the corresponding classifier. The "main" files, in either folder, are name Run*.py