Abstract: This study presents a machine learning approach for estimating the population of Melusi informal settlement in Pretoria using high-resolution drone imagery, LiDAR data, and open-source building footprints. A U-Net deep learning model was used to detect dwelling structures, while a hybrid ElasticNet and Random Forest regression model estimated population from dwelling counts. The system achieved accurate results, producing an estimated population of 41,329 people, close to the independently reported figure of 43,000. The research also included exploratory analysis of the imagery, LiDAR, and household survey data to assess data quality and limitations. Results show that combining drone imagery, Random Forest modelling, and machine learning can provide a faster and lower-cost alternative to traditional census methods for informal settlements.