Project for "Information retrieval" course - University of Milan - Year 2021/22
Authors: Marco Ghezzi - Matteo Limoncini
The aim of this project is to create an information retrieval system that allows to search and execute queries over a large dataset of medical images; the main purpose of the system is finding the images most similar to a given one. We applied two different machine learning techniques in order to build our retrieval system: supervised learning, using an artificial neural network to build a classification system, and unsupervised learning, in particular clustering. These different techniques were compared by analyzing different performance metrics and computational time. As a final result, we noted that employing a Convolutional Neural Network for classification (supervised learning) produces the best results in terms of precision and time efficiency