CloudWalk Technology, a Chongqing-based enterprise, won a double championship in the NIST-FRVT, the world's most authoritative facial recognition algorithm test, according to the company on March 15.
CloudWalk Technology won the championship of two items in NIST-FRVT.
The test, organized by the National Institute of Standards and Technology (NIST) of the US, is recognized as a definitive competition in facial recognition technology. The competition accepts algorithm submissions from facial algorithm vendors worldwide. After conducting confidential testing, NIST will publish the results and rankings on its official website. In the latest list, CloudWalk Technology ranked first in facial recognition algorithms for verification (1:1) and identification (1:N). It was also the runner-up in recognition of faces under masks.
NIST-FRVT, as the world's largest, most rigorous, and most competitive facial recognition algorithm competition, is known as the "golden benchmark" in the industry. With tens of billions of comparisons of different types of sample photos, it evaluates facial recognition algorithms to the accuracy of one in a million. Up to now, nearly 100 companies and research institutions have participated in this test, including VisionLabs, and Vocord of Russia, Ever AI of the US, and China's CloudWalk Technology, SenseTime, Megvii, Hikvision, Dahua Technology, Tencent, etc.
The good result of CloudWalk Technology in this competition is due to its algorithm design principles that fit the actual application scenarios. For example, in the face of the current COVID-19 epidemic, products launched by the company not only feature the recognition of faces with masks but also take into account the simultaneous presence of faces with and without masks. As the generalization ability improved, they can cope with more complex situations so that the algorithm can really work. The products' design principles enable the algorithm to show strong performance in lighting variations, sharpness, facial orientation, and races.