Video understanding is based on implementation of Neural Networks. These Networks are able to detect agreed upon concepts deriving from video content.
Today we can recognize behaviors, objects and understand them in relevant contexts. As an example, we are able to recognize an individual “wandering around” in a pre-defined area, such a behavior may be considered alarming in some areas’ such as an ATM machine, while considered normal in other areas, such as a bus station.
The Technology we use can automatically identify a movement type and a location type – with no manual setup. The combination of behavior and location results in a unique discernment of the event classifying it for human review within seconds from the event in Realtime.
Overcoming challenges in a two-folded approach:
The system uses NVIDIA GPU processors – providing for the high demand processing required for such calculations.
A unique architecture to shorten the processing time of each task, allowing the system to complete analysis in near real time.