viisights behavioral recognition solution can be deployed on-premise or via the cloud, and is scalable to an infinite number of cameras with data analysis provided within seconds of a triggered event. The innovative solution is based on real-time temporal and holistic video streaming analysis using video clips rather than standalone images, and employs convolutional neural networks and LSTM models for training its core AI engine. This creates a unique event signature that includes the scene participants and their extracted features, such as positioning, movement, size, and relationship with others. The information can be used to detect potentially dangerous or high-interest situations in progress, and to automatically send alerts to other investigative applications.
High-interest situations detected by the system include but are not limited to: people fighting, riots and vandalism, people brandishing weapons, blood on individuals, contextual loitering, a person abandoning an object, people wearing or not wearing masks, movements such as running, walking, or climbing, groups of people gathering or running, people climbing over fences or loitering near a fence, people entering a perimeter without permission, people in roadways, vehicles on sidewalks, vehicle collisions, traffic jams, people riding bicycles or motorcycles, people entering or exiting a vehicle, and smoke and fire detection.
viisights behavioral recognition solution successfully minimizes false alerts and their related costs while maximizing safety and security, providing a new level of video intelligence for predictive analysis, incident prevention, response management, and risk mitigation. By only analyzing general behavior patterns of individuals, groups, vehicles, and traffic-flows, viisights technology does not identify faces or license plates to protect personal privacy.