A research team has built a satellite-based pipeline that spots unconnected schools and nearby cell towers from orbit, no ground-level surveys required.
The system uses high-resolution satellite imagery and transfer learning to detect both schools and cell towers simultaneously. Trained on minimal labeled data, it adapts pre-trained object detection models to new geographic regions and then measures the spatial gap between each school and its nearest tower as a rough stand-in for connectivity. The team demonstrated the approach on real imagery from Lesotho, a landlocked country in southern Africa where ground data is sparse.
The practical target is the Giga Initiative, a UNICEF-ITU effort to connect every school on the planet to the internet. Mapping schools and their connectivity status at scale has historically depended on patchy third-party datasets; a vision-only pipeline that runs on satellite imagery sidesteps that dependency and could let Giga prioritize investment without waiting for governments to file accurate infrastructure reports.
The obvious caveat: detecting a cell tower near a school does not mean the school has a working connection or that the tower has capacity to spare. Proximity is a proxy, not a guarantee — and the gap between a dot on a map and a student loading a webpage is where most connectivity projects stall.