YITU takes top spot on NIST Facial Recognition Vendor Test leaderboard
YITU Technology has taken first place on the NIST facial recognition testing leaderboard for the second year in a row, beating out a pair of other Chinese organizations and a pair of Russian organizations, which rounded out the top five in the latest Face Recognition Vendor Test (FRVT) Ongoing.
An entry from the Shenzhen Institutes of Advanced Technology of the Chinese Academy of Sciences placed second for the lowest false non-match rate (FNMR) across three categories of Visa and mugshot photos. NTechLab and Vocord finished third and fourth, respectively, while Megvii came fifth.
FNMR rates for 1:1 algorithms were compared at the specified false match rates (FMRs) of 0.0001 and 0.000001 or lower for Visas, and 0.0001 or lower for mugshots.
YITU AI Research Scientist Dr. Wu Shuang, who is managing the company’s expanded research lab in Silicon Valley, told Biometric Update in March that the company’s technology also has less variable accuracy between different demographics than systems commonly cited in controversies around algorithmic bias.
IDEMIA had the top accuracy among algorithms for 1:1 matching in the selfie and webcam categories in NIST FRVT tests earlier this year.