Rank One Computing (ROC) has shown a dramatic improvement in its most recent NIST FRVT Ongoing benchmark evaluation. The company submitted the 2.0 version of its facial recognition SDK, which proved to be 200 percent more accurate than prior versions of the platform.
That represents significant progress considering that ROC’s algorithm was last tested only five short months ago. The company noted that its accuracy now rivals that of the world leaders, and is now within half a percentage point of the most accurate solution in every single category in the FRVT test. In some cases, the difference between that solution and ROC’s offering was only 0.1 percent.
ROC 2.0’s average accuracy across all categories is 99.4 percent. The company also says that its algorithm was much more efficient than the other top competitors, and believes that that should make its offering more appealing to potential clients. The facial recognition SDK can be used to help with identity verification for organizations that need to comply with Know Your Customer regulations, or with deduplication efforts for those with large databases.
The latest version of ROC also introduces a few new video analytic features, including liveness detection, object detection, age verification, and optical character recognition. Object detection can be used for people and vehicle tracking, while optical character recognition allows the platform to read physical credentials and identify license plates.
“The accuracy of this algorithm is representative of how powerful ROC’s face recognition has become on diverse imagery, with accuracy measured on persons from around the world,” said ROC CEO Scott Swann. “With these substantial error rate reductions, ROC’s AI/ML SDK 2.0 has effectively reached parity with the highest ranked FRVT performers”.
As an American developer, Rank One is hoping that its solution will appeal to US government agencies and other organizations that are hoping to partner with a domestic facial recognition provider. The company has repeatedly warned that products developed outside of the country pose a national security threat to the US.
February 4, 2022 – by Eric Weiss