Students of the Department of Artificial Intelligence Systems, Lviv Polytechnic National University, have created a prototype of a security system that distinguishes a living person from a photo or video image. The built-in algorithm identifies fakes in real time. According to the developers, the technology can be used to equip security systems at the front door, identify suspects in crimes or search for missing persons. The work lasted 5 months, faculty of the Department of Artificial Intelligence Systems and mentors of the international partner company GlobalLogic joined its implementation.
Third-year students from Lviv together with experienced engineers designed a high-precision identity verification system. The algorithm in real time identifies the face and checks if the person is logged in personally or it is a printout, photo or video. If the technology determines that the image is fake, it will signal an error or attempted unauthorized entry.
The system analyzes several frames in a short period of time. The algorithm takes into account the possibility of distortion of the human face during identification, different environments and similarity of facial features. With the help of artificial intelligence, the system notices even a slight difference between the image of a person from a real person, making it impossible to pass the test with someone else’s photo. The process of analyzing a person is due to a neural network that creates a unique set of values for each person. The process of facial analysis is similar to the technology of user recognition in the iPhone.
Experts from the international IT company GlobalLogic were the students’ mentors in the process of developing. They helped to determine the technological component of the project, solve complex problems in the process, systematically guide their work, teached to work in a team and advised on related issues.