Apprendimento federato per la classificazione di immagini aeree

In remote sensing technology, data privacy is a relevant issue in aerial imagery, due to the sensitive nature of geographical data which often includes critical infrastructure and private property details. This thesis addresses this challenge by implementing Federated Learning (FL) techniques in the classification of aerial images. This project's major contributions include (i) the application of transfer learning to extract data features from an image, which aims to enable careful pre-processing and data enhancement for data training purposes; (ii) the application of FedAvg and FedProx algorithms together with pre-trained CNNs in a FL framework to ensure customer privacy while maintain good performance; (iii) the integration of a customized dataset into a FL framework.