Rank One is boasting about its strong performance in the latest round of NIST testing. The company’s ROC SDK v1.20 was one of the most accurate algorithms tested in the most recent Face Recognition Vendor Test (FRVT) 1:N identification report, which evaluates a solution’s ability to match an image of an unknown person’s face to an identity in a database.
Based on the results, Rank One argues that its combination of accuracy and efficiency sets the ROC SDK apart from its competitors. Though the algorithms from NEC and Paravision were slightly more accurate, Rank One’s solution was considerably faster. Its biometric templates have a much smaller data footprint, and Rank One was able to generate those templates (and carry out matches) far more quickly.
That last bit is important because it can dramatically reduce the hardware requirements for companies that are looking to implement a facial recognition system. Some of the applications of the technology include mobile identification and surveillance, where templates are generated from real-time video feeds.
Rank One also notes that templates need to be updated regularly as facial recognition technology continues to improve. That process inevitably takes less time with a smaller template, and that can create more value for an organization that needs to maintain a larger database. With that in mind, Rank One indicated that it will be releasing a new version of its ROC SDK within the next few weeks, and that its customers will not need to purchase a new license to take advantage of the promised accuracy gains.
The NIST results come in the wake of an Adroit Market Research report that predicted that the facial recognition market will reach $12 billion by 2025. Grand View similarly expects facial recognition to be a key component of a $110 billion image recognition market.
March 26, 2020 – by Eric Weiss