The project is part of Facebook’s Connectivity Labs, the technical arm of its Internet.org initiative that deals with drones, satellites, and lasers for delivering internet to rural areas and developing countries. With better maps, the company is able to determine whether Wi-Fi hotspots or cellular technologies are better for bringing people online — and helping them sign up for Facebook naturally.
To generate the maps, Connectivity Labs worked with Facebook’s data science division, infrastructure unit, and machine learning and artificial intelligence groups. The coalition analyzed satellite imagery for 20 countries covering 21.6 million square kilometers, which amounted to 350TB of data. Using a mixture of computer vision techniques, including the image-recognition engine Facebook uses to identify people’s faces in photos, the team was able to identity human-built structures. The company stresses it didn’t use any Facebook photos to create its data.
Facebook honed the maps by applying machine learning and AI techniques, which “succeeded in identifying outlines of buildings and highlighted those for which it had high confidence while suppressing areas not likely to contain human-made structures.” Once the structures were laid out, Facebook was able to use them as proxies for where and how people live. Using census data, the team redistributed the population data sets evenly across each location, under the assumption that the method was the least error-prone way of determining how many people lived in each building.
The company says it will be releasing the data to the general public later this year. “We believe this data has many more impactful applications, such as socio-economic research and risk assessment for natural disasters,” Facebook said in a statement. The team has plans to work with Columbia University’s Center for International Earth Science Information Network to create a new, combined population dataset later this year as well.