Identifying gaps in OpenStreetMap coverage through machine learning
“Having found our model to work sufficiently well, we apply it to predict OSM building footprint for all of Haiti, and flag cells that are predicted to be fully mapped but are actually not covered by OSM .”