AgentVI

The core of Agent Vi’s technology is its distributed architecture, which distributes the video processing task between an edge-device and a server, enabling optimal utilization of the processing elements at each end, reducing bandwidth between both ends as well as hardware costs. Agent Vi’s unique deep-learning implementation leads to the lowest ratio of hardware cost per camera, compared to other software solutions. This architecture can be implemented in the public cloud as well as in private networks. In addition, Agent Vi™ has developed its own object detection and classification models, enabling distinction between different classes of objects (people, classes of vehicles etc.). These models have been trained on actual surveillance footage in a variety of environments and weather conditions, resulting in accurate and robust object classification in the most challenging environments. These models are continuously enhanced and improved, and new models are added. On top of its classification model, Agent Vi™ developed a broad set of video analytics functionalities, including rule-based real-time event detection, autonomous anomaly detection, video forensics and statistical analysis under a single software platform. All functionalities rely on a single video analytics processing pipeline and can be implemented simultaneously on any camera without material impact on the required hardware resources.