The social network says that it is creating the most accurate maps of population density, so is analyzing 14,600 million images captured from satellite. For the better of everyone, they plan to make public these works, and today teaches us something of what has advanced.
Using algorithms of artificial intelligence and systems of machine learning, Facebook is recognizing and assembling information that is removed from each of the images. The project is in charge of the Connectivity Lab and it has been announced today at the Mobile World Congress in Barcelona.
Knowing where more people live they can better deploy its services
Real Facebook intends to learn more about the areas with large agglomerations of population, once identified, the deployment of its drones unmanned will be more efficient.
These drones are known as Aquila and work with solar energy, its mission is the connectivity to remote points of the planet. In the picture below you have the appearance that looks one of them, they are large and able to plan for months over the area:
We speak of recognizing structures where people live in more than 21.6 million square kilometers
Returning to the topic of density maps, are being collected billions of images from satellite, and the way of identify buildings and infrastructures where people live, is using artificial intelligence.
In areas with census, values have contrasted to verify that the error level is quite low.
Right now the study is underway on twenty countries, almost everyone in developing, with large areas without people – is easier to identify the nuclei-: Algeria, Burkina Faso, Cameroon, Egypt, Ethiopia, Ghana, Côte d’Ivoire, Kenya, Madagascar, Mexico, Mozambique, Nigeria, South Africa, Sri Lanka, Tanzania, Turkey, Uganda, Ukraine and Uzbekistan.
That said, the information has many utilities, but Facebook wants it to position best drones, satellites and services. Fortunately institutions, companies, Governments, and organizations can benefit from it.
On the right the result. This paper develops mainly two things: the work of computers with millions of images and the recognition of structures. Below an example: