AnyVision has optimized its Recognition AI software for Ambarella’s CVflow System-on-Chip (SoC) solutions. The improved software has a lower computational footprint, and will allow organizations to run various computer vision applications directly on the camera itself.
In plain terms, that means that organizations will not need to maintain as many on-premises servers to process all of the information gathered through their security cameras. Those cameras will be instead be able to perform facial recognition and other surveillance functions at the edge to minimize power and infrastructure costs for AnyVision customers.
Potential applications of the Recognition software include biometric access control in addition to more traditional surveillance. AnyVision’s facial recognition tech can identify registered faces with 99 percent accuracy, and offers liveness detection to make sure that those faces belong to real people (as opposed to a photograph that might be used in a spoofing attempt). Since the technology is touchless, recognized individuals will not need to make physical contact with a terminal or a scanner in order to gain access to a building.
On the surveillance front, AnyVision’s tech supports the creation of watch lists, and can notify administrators whenever it spots someone on one of those lists, even if that person is only one face amongst a larger crowd. The software also enables person detection and analytic utilities, and can function in low light and extreme angle situations.
The updated AnyVision software is specifically geared toward the Ambarella CV2x family of edge SoCs. Meanwhile, Ambarella’s open SDK will make it easier for customers to integrate other third-party applications into the AnyVision platform.
“Because of their high data volumes, compute demands and latency requirements, cameras are the most challenging ‘things’ in the Internet of Things,” said AnyVision’s CEO, Avi Golan. “By running our video analytics and Recognition AI on Ambarella’s CVflow SoCs, we’re pioneering new frontiers in edge AIoT computing, which opens up whole new use cases and markets.”
Paravision added support for the CVflow platform back in December. Ambarella has since partnered with Lumentum and ON Semiconductor to release two new reference designs for AIoT development, while AnyVision has encouraged the NIST to draft guidelines that can inform the creation of unbiased facial recognition systems.
June 22, 2021 – by Eric Weiss