The Liza Alert search team has begun testing a technology that allows for searching for missing people. The self-learning neural network will analyze photographs taken by drones and look for people on them.
The developer of this neural network is VimpelCom telecommunications company; the network is currently being taught to differentiate photos featuring people and their belongings from the ones where there are no people on them, Izvestia daily reports. Both staged and real photos taken by Liza Alert are being used.
According to Izvestia, volunteers used to handle all photos manually. It took a group of 30 people at least 5 hours to review 2.5K images taken by Liza Alert drones. Neural networks require much less time: some 2.5 hours, and their accuracy reaches 98%.
This June alone, Liza Alert received almost 2,000 missing people requests. The organization found 1,662 of the missing persons, with 1,526 of them being alive. The movement members say that the use of artificial intelligence for image analysis will be especially efficient if the photos are not taken in the cities where too many people could be captured on a photo.