The Skolkovo Foundation resident company VisionLabs has for the third time in a row confirmed its reputation as one of the global leaders in the field of computer vision and machine learning. This is based on the results of the independent testing of facial recognition algorithms by the National Institute of Standards and Technology (NIST) in the United States.
The NIST ranking is the most authoritative source on the efficacy of global recognition systems. Solutions are tested on a multimillion closed database of images, allowing for objective results. The National Institute of Standards and Technology was founded in 1901 and is now a part of the Department of Commerce of the United States. Every year, the institute conducts voluntary independent certification of technology companies from across the world. In this way, the best developments from different fields of science and technology are determined, including information technology related to biometric research and facial recognition in particular.
Alexander Khanin, the founder and chairman of VisionLabs. Photo: Sk.ru.
Testing takes place on a number of image types, from standard document photos to pictures from web cameras or phones. The main goal is to determine the most universal facial recognition algorithms that work equally well in different application scenarios.
VisionLabs’ solution was the best in the following categories:
“Our team is made up of leading researchers, which allows us to constantly develop and improve the company’s products,” said Alexander Khanin, the founder and chairman of VisionLabs. “This year the entire world was faced with new challenges to develop solutions for the fight against the spread of the coronavirus. We are proud that VisionLabs’ complex biometric products are also important instruments for protecting people’s health and safety.”
This is in reference to the company’s software and hardware complex, VisionLabs THERMO, which is designed to measure and monitor a person’s body temperature contactlessly.
Over 140 of leading companies from across the world took part in NIST testing this time round. According to the results, the speed of VisionLabs recognition algorithms is one and a half times greater than others presented among the top rated. Aside from that, the high optimization of neural network architecture and efficient storage of biometric templates means that the company’s solutions need four times less memory, which is of great benefit to the client. This is particularly important when working with large databases at banks, transport and industrial enterprises – fields where VisionLabs developments are being implemented the most.
“Unlike 99% of our competitors, we work in the commercial sector and we periodically have large-scale ideas,” said Alexander Khanin in a recent interview with Sk.ru.
The tests proved that the VisionLabs algorithms work efficiently with photographs from any racial group and can be applied successfully around the world. For example, for the Visa dataset, which contained images of people of all ages from over a hundred countries, the accuracy was 99.75% with a false probability of one in a million.
“We are proud of another success of a Skolkovo Innovation Center resident. For a third time in a row the VisionLabs facial recognition algorithms have taken first place in the prestigious rating of the National Institute of Standards and Technology of the United States,” said the chairman of the Skolkovo Foundation Arkady Dvorkovich. “For winning this award, I sincerely congratulate the company’s team of high-class, dedicated specialists and talented scientists who are creating breakthrough, in-demand products and technologies. Of course, such victories are significant for the development of domestic innovations.”
In previous NIST tests, VisionLabs took first place for the Mugshot category as well as being in the top three in the Visa category. The company’s algorithms were between two and two and a half times faster than all the main competitors. In March of this year, VisionLabs came first at the largest international competition – ChaLearn Face Anti-spoofing Attack Detection Challenge – for the second time in a row during the CVPR 2020 conference, the main annual computer vision event. Over 340 scientific teams took part, including Intel, Hisign and others.