AnyVision is boasting about its strong performance in the NIST’s most recent 1:1 Face Recognition Vendor Test (FRVT). The company’s algorithm made it into the top 10 in the overall rankings, and was the top-ranked solution from a developer in the western world.
The FRVT looked at 177 different algorithms, and evaluated each in five performance categories. AnyVision did well across the board, recording good scores for matching accuracy and matching speed on its way to the finish. The company’s solution also did well with regards to storage, memory consumption, and resilience.
AnyVision noted that its technology was not designed to match static images like those used in the NIST test. The algorithm was instead built to identify faces in video footage, and to run well on low-power embedded devices. AnyVision argued that that is a more difficult task, and the fact that it did so well at a different task speaks to the robustness of its technology.
“These results reflect the significant investments AnyVision has made to ensure that the algorithms that power its AI engine are superior in the market,” said AnyVision CEO Avi Golan. “We believe that AnyVision is the AI for IoT. We have taken the crucial first steps towards bridging the digital-to-physical divide, with a platform for understanding the way people experience and interact with the physical world.”
AnyVision’s facial recognition tech has previously been used for surveillance and access control, though the company has placed a stronger emphasis on the latter in the past few months. The company launched a new Touchless Access Control solution for physical venues in July, and later secured $43 million in funding to help scale its operations.
More recently, AnyVision detailed a set of ethical considerations to encourage the development of unbiased facial recognition systems. In doing so, it called particular attention to the need for diverse data sets when training facial recognition algorithms.
March 12, 2021 – by Eric Weiss