Top performing developers steady in updated NIST facial recognition 1:N test results
The latest edition of the U.S. National Institute of Standard and Technology’s (NIST’s) Face Recognition Vendor Test (FRVT) 1:N report, dated April 28, shows continued improvement among newly-submitted algorithms not just in biometric accuracy, but also size and speed.
Algorithms from SenseTime and CloudWalk sit atop the overall leaderboard, followed by Paravision and Idemia. The lowest error rate in each category was also recorded by one of the above four, except for the Border at 10 or more years category, which an NEC algorithm had the lowest error rate for.
Hangzhuo AIlu Network Information Technology is the only developer that submitted a facial recognition algorithm for the first time, but HyperVerge, Qnap Security and RealNetworks (which makes SAFR) also returned to the benchmark.
Following a February change to the FRVT API, NIST shows false negative identification rates (FNIR) in blue for algorithms detecting faces and producing templates from multiple faces in a single image, which the Idemia and HyperVerge algorithms are examples of.
Market forecast to reach $22.5B by 2032
The global market for facial recognition is forecast by Future Market Insights to grow from $5.2 billion this year to $22.5 billion by 2032, a compound annual growth rate of 15.7 percent. Emerging security camera technologies and deployments in public spaces by governments are expected to lead the growth.
Other areas of technical advances expected to drive growth include 3D imaging and cloud-based solutions.
Related technologies like face detection for intelligent signage are also expected to register significant gains.
Biometric software is expected to grow at an even faster rate than the overall facial recognition market, as are law enforcement applications.