AnyVision’s Recognition AI Software Running on Ambarella’s CVflow® SoCs
AnyVision has announced the availability of new AI software designed to run on Ambarella’s CV2x family of edge AI vision SoCs, and helps make cameras smarter by extracting valuable data from high-resolution video streams in real-time.
The new AnyVision Recognition AI software is available now and optimized for Ambarella’s open and mature software development kit (SDK), enabling low-power, high-resolution IP security cameras with features such as facial recognition, person detection, analytics and more. This software, in combination with Ambarella’s highly efficient CVflow® AI vision SoC architecture, allows cameras to perform complex data analysis and deliver breathtaking imagery, while preserving vital system resources such as power and network bandwidth.
Historically, organizations that wanted to add video analytics and intelligence to their IP security cameras or turnstiles relied on expensive on-premises servers, which were required to handle the compute loads. Ambarella’s edge AI vision SoCs push more of this processing to the actual cameras and dramatically reduce the need for on-premise servers, fundamentally changing the historical paradigm for large-scale video security and biometric-based access control.
“Because of their high data volumes, compute demands and latency requirements, cameras are the most challenging ‘things’ in the Internet of Things,” said Avi Golan, CEO at AnyVision. “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.”
The open SDK for Ambarella’s CVflow SoCs allows easy integration of third-party applications, while enabling OEMs to address several emerging use cases, including:
Touchless access control. In order to minimize physical contact between users and surfaces, new touchless access control solutions are evolving that capitalize on facial recognition. Running on Ambarella’s CV2x series SoCs, AnyVision’s Recognition AI software uses deep neural networks to extract and compare the faces of those entering a facility with those registered in the system, providing a recognition accuracy rate of approximately 99%.
Video surveillance. Instead of relying on expensive on-premise servers to process and interpret the video streams from low-power, high-resolution security cameras, the cameras themselves can become powerful edge computing devices with embedded facial recognition and watchlist-alert functionality. AnyVision’s AI software can overcome the most challenging conditions — from large crowds to low light environments, extreme angles, and obscured faces.
Liveness detection. This edge AIoT technology helps organizations take advantage of 2D and high-resolution 3D liveness detection, which is becoming increasingly important as criminals use spoofs (e.g., a 2D picture held up to the camera by an unauthorized person) in order to bypass the access control or video surveillance system.
“The ability to perform real-time video analytics directly on cameras is a growing trend in the convergence of AI and IoT,” said Fermi Wang, president and CEO of Ambarella. “The combination of AnyVision’s Recognition AI software and our CVflow edge AI vision SoCs will enable smarter AIoT devices for the next generation of security and access control systems.”
This partnership empowers organizations of all stripes to harness the power of edge computing to analyze traffic, security, crime and liveness in real-time — and in a much more efficient and cost-effective manner.