Mapping the world through machine learning

Mapping the world through machine learning: Facebook researchers are using machine learning models on high-resolution satellite imagery to produce detailed maps showing human settlement around the world, including this image of North Egypt, courtesy of Visual Capitalist. With 95% of our population living along the banks of the Nile, Egypt is a highly illustrative example of the power, detail and beauty of this kind of sophisticated mapping, which has been made possible with machine learning. If you look at the full-resolution version of the map, you will see that even individual buildings are visible, with human settlement depicted on a granular scale. The maps, believed to have an accuracy rate of 99.6%, don’t simply provide interesting data; they could also be used to improve living conditions — particularly in developing and rural areas.