Generative adversarial networks : virtual satellite images generation

Details

Supervisors

Schaus, Pierre

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique

Abstract

This works studies the performance of conditional generative adversarial networks in the context of satellite imagery. Using different losses function and network architectures, several models are tested and assessed, with the purpose of performing efficient data augmentation and easing the development of new satellite-based applications, a domain where acquiring labeled data is expensive.