This repository contains a project developed as part of Digital Signal and Image Management course of the Master's degree in Data Science, University of Milano Bicocca.
This project aims to apply three digital signal and image management techniques:
-
Mono-Dimensional signal classification: using cough recordings as input data, the objective is to develop a multi-class classification model capable of classifying recordings by subject’s status with comparable or better results than expert physician.
-
Bi-Dimensional signal classification: using x-ray brain tumor images as input data, the objective is to develop a multi-class classification model capable of classifying x-ray images and potentially detect different types of brain tumor.
-
DCGAN: using a dataset containing celebrity faces, the objective is to develop a Generative Adversarial Network (GAN), that can generate completely new images that do not exist by learning from the input data with the implementation of neural networks. In this case study, a Deep Convolutional GAN has been developed, which can better capture local structures in images.
This repository is structured as follows:
.
├── DSIM Project
│ ├── DCGAN - Roberto Ferrari
│ │ └── DCGAN.ipynb # DCGAN task notebook
│ ├── DSIM presentation.pdf
│ ├── bi_dim_signal - Doghmi Samir # Bi-Dimensional signal notebook
│ │ └── task_2.ipynb
│ └── mono_dim_signal - Luca Iarocci # Mono-Dimensional signal notebook
│ └── COUGHVID_classification.ipynb
└── README.md