Synesis has one of the world’s top 10 facial recognition algorithms
Synesis ranked #9 in NIST’s new Border test, which assesses facial recognition in non-professional frontal photographs taken under imperfect conditions.
NIST – U.S. National Institute of Standards and Technology. Using its own tests, the Institute compiles a rating of developers of face recognition algorithms and updates it on a regular basis. Companies can submit multiple applications, but not more often than once in 4 months. The latest results of the competition were published at the end of July, in which 224 solutions from 140 developers were evaluated.
In tests, algorithms compare facial photographs and identify pairs that depict the same person. The test determines false identifications when the algorithm sees different people as one person, as well as false skips when the program is unable to identify one person in different photos. Algorithms are ranked according to the level of such errors – the fewer there are, the higher is the ranking.
In the Border test, Synesis’s solution showed an error rate of 1.05%, which is just slightly different from the best result (0.79%).
Vadim Petrov, the developer of Synesis Computer Vision lab: “The peculiarity of the Border test is that the algorithm recognizes and compares frontal photographs taken in arbitrary conditions and with non-professional cameras, similar to those taken at the state borders and customs. Faces in such pictures are often not centered and can be positioned slightly higher or lower of the frame center; lighting is also imperfect on such pictures. However, the solution shows high efficiency not only in identifying people at the border, but also, for example, at the clock in&out stations at the entrances to work areas, as well as in any other situations where the camera takes a face photo.”
It’s the second year for Synesis participating in the NIST competition. The company will send the next application for the competition in October. By that time, the team plans to implement the latest developments in the field of face recognition and develop new algorithms.
Synesis uses facial recognition technology in its Kipod Smart City platform. With the help of video analytics modules, Kipod detects traffic violations, allows you to find stolen cars, search for criminals and missing people, and respond to events such as smoke or gunfire in real-time. The platform is scalable for unlimited users, cameras, and archive depth. Today Kipod is using in 5 countries all over the world – in Ireland, USA, Russia and Azerbaijan.