NtechLab reaches leaderboard in less constrained categories of NIST facial recognition test

The National Institute of Standards and Technology (NIST) has released the report for its Ongoing Facial Recognition Vendor Test (FRVT), Part 1: Verification, which shows NtechLab among leaders in technology for the most complicated image sets, according to a company announcement.

An NtechLab algorithm ranked second in testing for the CHILD EXP dataset, and another ranked seventh for the WILD dataset, as the company says it doubled its accuracy compared to the previous test. These categories cover less constrained conditions, and the leading algorithms in them are used for public security purposes, according to the announcement. The same two NtechLab algorithms also finished in the top 10 in the VISA 0.0001 category.

“Several new companies, including Chinese firms, participated in this NIST testing. We can say that competition was never as tough as this time,” says NtechLab founder and head of the company’s neural networks laboratory Artyom Kukharenko. “Now we can happily state that regular improvements of our algorithm help to maintain its international leadership in accuracy of face recognition in various conditions.”

More than 100 algorithms were tested by NIST for the 1:1 portion of the FRVT.

NtechLab won first place in two of three categories of the Face Recognition Prize Challenge (FRPC) hosted by IARPA in 2017, and came second in the other. Moscow police are piloting AR glasses with facial recognition from NtechLab, and the company recently formed a partnership to apply its FindFace algorithm to retail intelligence.

Article Topics

accuracy  |  biometrics  |  facial recognition  |  NIST  |  NTechlab  |  testing