VisionLabs, part of Sberbank ecosystem and one of the global leaders in computer vision and machine learning, for the second year in a row won in ChaLearn Face Anti-Spoofing Attack Detection Challenge international competition, Sberbank’s website reports.
More than 340 companies and research teams from different countries, including Intel, Hisign, etc., took part in the competition. VisionLabs presented Liveness solution, which was twice as good as the second runner-up.
The competition’s main objective is to recognize if there is a real human in front of the camera, rather than a fake as a printed photo, video or a mannequin. This year, the organizers made the terms more tough compared to last year. The competition was held to identify Liveness attacks using three databases with photos of people of different races. One database was given for learning, and the other two for testing. Participants’ algorithms were evaluated on attacks that had not been used during the learning and testing stages and which are critical in various face recognition areas, especially in the financial industry. The contending algorithms should determine whether a person facing a camera is a real one, or an image on a phone/tablet screen, or a 3D mask.