Bear Grylls and CSI detectives had years of practice in following traces, but these drones learned to do their job much quicker. Swiss researchers worked on an AI capable of making drones almost as perceptive as humans at deciphering footprints.
Small quadcopters learned to follow man-made trails in forests as a way of supporting rescue teams in search of lost hikers. According to project leaders, about 1000 people go missing each year in Switzerland mountains alone! This is why they started developing a software that uses a Deep Neural Network (a program capable of learning from errors to rectify and improve next experiences) to identify never-before-seen human trails.
Researchers did most of the heavy work themselves, snapping over 20.000 images of trails using camera headsets and feeding the data to the machine. Before you know it, a small quadrotor with just two tiny cameras began pursuit with the AI algorithm on “deck”. Remarkably, the UAV followed the right trail 85% of the time, with 3 percent more than humans.
The difference is not that noticeable, granted, but the idea that an artificial intelligence program embedded in a mass-market drone can replace efficiently human trackers is nothing short of incredible.