Population estimation utilizing Earth Observation data

Abstract

A thorough understanding of population distribution could aid the government in many decision-making processes. This thesis promotes the creation of up-to-date and detailed population maps by curating a large-scale data set for population estimation and, second, developing a deep learning-based framework to infer the population count/density and improve the model's transparency using an explainable AI technique. Finally, it generates high-resolution population maps at the building level.