Facebook is using AI to map the world’s population density. Earlier in 2016, the company already unveiled the concept by creating maps for 22 countries. Currently, it expanded with new maps that cover most of the African sub-continent.
In a blog post, the social media giant said that the project will ultimately map nearly the entire world’s population. As per explanation by Facebook, it’s not an easy task for human beings to create maps. The high-resolution satellite imagery that the company own covers the majority of the places around the world but the process of converting it into useful information would require a lot of time.
Population density maps can be created by labelling each building in the images and after that, the process needs to be cross-referenced with census data. In the case of Africa, the process is a bit complicated census tracts can cover a region of 150,000 square miles but track just 55,000 people.
Fortunately, this kind of work is very tedious but simple, which is perfect for AI. The company’s engineers used these steps to automate the process, they used data from open source mapping project Open Street to train a computer vision system that has the potential to identify building in satellite imagery. After that, the engineers used the technique to eliminate most of the data produced by the satellite. The data were based on unoccupied land.
According to the company, the new ML systems are faster and more precise than those unveiled in 2016. The new mapping work was verified by some researchers from Columbia University. The researchers were from the Center for International Earth Science Information Network (CIESIN).