ROC Pushes Biometric Presentation Attack Detection Accuracy Beyond 99% in SDK Upgrade

“The enhanced SDK also has a new standalone Facial Analytics algorithm that can estimate dozens of facial features in under 50 milliseconds on a single CPU core.”

Rank One Computing (ROC) has delivered a number of technological improvements in an upgraded Software Development Kit for facial recognition and computer vision.

Version 2.3 of the ROC SDK has a much more accurate single-frame passive liveness detection system, with an error rate that is two to five times lower across different spoofing mediums. Rank One says that the improvement brings its Presentation Attack Detection accuracy up to over 99 percent.

The enhanced SDK also has a new standalone Facial Analytics algorithm that can estimate dozens of facial features in under 50 milliseconds on a single CPU core. The speed has been delivered even while maintaining features like mask detection, age estimation, and other analytics.

The platform’s improved tattoo recognition algorithm, meanwhile, has error rates that are an order of magnitude lower than the last published benchmark from the NIST Tattoo Recognition Technology Evaluation. And its license plate detection capabilities have been improved across a range of conditions.

The upgrade comes after Rank One’s announcement of an important new strategic partnership with Anonybit, the startup behind a pioneering decentralized identity solution, in August. That announcement was quickly followed by the news that the Marion County Board of Education in West Virginia had opted to use ROC’s facial recognition technology for school security.

–

November 11, 2022 – by Alex Perala