Here you can find our news
We are glad to share the news about our first participation in the independent rating of face recognition algorithms (FRVT), compiled by the US National Institute of Standards and Technology (NIST).
To participate in this testing, we prepared one of our fastest facial recognition models for embedded systems. Such niche models are used by our customers to work on board low-power devices (video cameras, intercoms, controllers, etc.).
On the reports provided by NIST, you can notice that our system consumes many times less computer resources than the products of competitors. At the same time, our algorithm provides the minimum necessary quality metrics to solve a wide range of access control or identity verification tasks. To be more precise, our product allows one false-negative response per 50 persons (FNMR = 0.02) with a fixed threshold of false-positive response within once per 10,000 persons (FMR = 0.0001).
Moreover, according to the test results, we can say that not a single algorithm in the rating is capable of solving the verification task at the same time faster and more accurately than our algorithm.
Our next step will be to prepare and send for testing a more “heavy” model of persons verification with increased accuracy.