Giga's AI Satellite Mapping Reaches Every School in São Tomé and Príncipe

São Tomé and Príncipe, a small Central African island nation, lacked a complete, reliable inventory of school locations, which made it difficult for the Ministry of Education to plan and finance internet connectivity for classrooms.
Giga — the joint UNICEF–ITU initiative — applies machine-learning models trained to recognise school buildings (playgrounds, distinctive roof shapes) in high-resolution satellite imagery, then cross-references candidate sites against government census data, OpenStreetMap and Overture Maps before local staff verify flagged locations on the ground. IEEE Spectrum reported in December 2024 that this weakly-supervised deep-learning pipeline, combining transformer and CNN-based models, had by then mapped roughly a third of the world's schools across 141 countries.
By February 2025, Giga reported that 100% of São Tomé and Príncipe's basic and secondary schools had been mapped and validated, after which infrastructure modelling (cell towers, fibre routes) was used to give policymakers the tools to plan and finance connectivity.
Public documentation does not disaggregate accuracy figures or costs specifically for São Tomé and Príncipe, and the practice addresses the infrastructure and administrative precursor to connectivity — not classroom teaching or learning outcomes directly.

Read the full analysis: https://spectrum.ieee.org/school-internet-connectivity-globally-giga