N-Tech.Lab became the leader of the world challenge in face recognition algorithms, organized by The University of Washington. The algorithm created by Russian programmers upstaged more than 100 other systems, including the one called Facenet created by Google, as leading Russian business daily Vedomosti reported.
MOSCOW, Dec. 16, 2015 /CNW/ - N-Tech.Lab team developed a special face recognition algorithm based on the artificial neural network (a family of models inspired by biological neural networks), which allows to extract characteristic features of a person's face. N-Tech.Lab's solution was declared the most accurate one applying for the search in large datasets. They achieved 73.3% accuracy on 1 million faces.
In this nomination, Google's solution Facenet and the algorithm proposed by the specialists of Beijing University were the second and the third best ones respectively.
The biggest challenge in developing such a recognition technology lies in making the algorithm detect key features regardless of filming and lighting conditions, facial expression and image quality, the company representative noted. The software also has to crunch through large amounts of imaging data in a short time and with limited computational resources.
"Currently we are planning to launch a couple of commercial products based on our technology and we also consider the possibility of raising investment for future development" - Artem Kukharenko, N-Tech.Lab's founder, said.
Face recognition technology can be used by law enforcement bodies to search for and identify possible or wanted criminals using streaming video cameras, as Artem Kukharenko specified. Furthermore, this kind of information might be helpful for customs service in distinguishing people who are forbidden to enter a country. It could also automatically verify fans and monitor the situation on stands and surrounding areas during mass sport and culture events in a real-time mode.
"This kind of technology is currently successfully used in business-analytics systems, where they help to prevent bank frauds. The neural networks have recently "exploded" and nowadays they allow us to solve the tasks that used to seem too difficult. Their value and functional area will significantly increase in future", - the developer says.
Another field of the developed technology's application is direct marketing. The algorithm is able to analyze recordings from e-shops and search for the customers in social networks. Then the prospective buyer could receive personalized ads mails. At the moment N-Tech.Lab is negotiating with several large retailers about developing such solutions.