Integrating High-Resolution Imagery and Machine Learning for Small and Informal-Area Population Estimation in Melusi, Atteridgeville

Thabo Chesane , Kgoboko Maripane

Partner: ggm

Year: 2026

Abstract: This project aims to develop a methodological approach for estimating small-area population in Melusi (an informal settlement near Atteridgeville, Pretoria West) using high-resolution aerial imagery and machine learning techniques. This project will apply a fine-tuned deep learning model to extract building footprints from drone imagery and use building density metrics together with household survey data to construct a Bayesian regression model for estimating population size and density.

Presentation Video