<p dir="ltr"><b>This dataset contains the first nationwide map of building functions for Bangladesh. It classifies ~10 million structures using a robust, scalable semi-supervised learning (SSL) approach based on Random Forest. The model leverages OpenStreetMap for labels and remote sensing (VIIRS nighttime lights, GHSL, COP-DEM30) for features, providing a valuable resource for urban planning, disaster risk assessment, and sustainable development policy.</b></p>
