Contemporary machine learning can dramatically improve mapping efforts by narrowing down the search space for points of interests (POIs). The trained model can be run anytime to update the predictions frequently. An accurate school map, produced by a human-in-the-loop type of approach (combining machine learning and the expert mappers together), will still need field agents on the ground to further validate and confirm ‘unrecognized’ schools with local authorities.

Our experience here is confirmed by other experiments. Recently an AI-assisted high-voltage grid mapping model developed by Development Seed has been successful in rapidly mapping power infrastructure. In this case, a similar model and workflow was used to detect high-voltage towers in Pakistan, Nigeria and Zambia that the expert mappers afterwards used to map the high-voltage grid at the country-wide scale. The grid map now lives in OpenStreetMap that the platform users who have local knowledge about the electric grid can update and correct the grid. Machine learning has provided a means to study the SDGs related issues the school mapping case we present in this project address it’s another example of machine learning or AI assisted infrastructure mapping from satellite imagery.

Thank you for your interest. If you are interested in deriving insights from satellite imagery and AI-assisted POIs mapping, we are happy to work with you.