Joint project for "Algorithms for massive datasets" and "Statistical Methods for Machine Learning" courses - University of Milan - Year 2020/21
Authors: Marco Ghezzi - Matteo Limoncini
This project analyses and compares different solutions to tackle the problem of image classification for the Glasses or No Glasses dataset. We start from a simple neural network up to more complex Convolutional Neural Networks with multiple layers. The goal of this project is to implement an artificial neural network in order to identify if a person is wearing glasses or not. We used the TensorFlow API to build and train our models and the Keras Tuner framework for a better hyperparameters optimization. This project identified the best suited model to solve this problem: a convolutional neural network with 3 convolutional layers and 2 dense layers that reach an accuracy of 0.99 and a loss of 0.10