Scalability
In real-world deployments where many video streams are processed it’s critical to optimize the use of the CPU and GPU. Customers and system integrators want to run EyesOnIt on a server that will process all their video streams but still fit within their budget. The EyesOnIt model that compares your object descriptions to your video stream is referred to as a Large Vision Model. It is a large and highly advanced model. It consumes a lot of computer resources to compare object descriptions to each frame of your video. To optimize your CPU and GPU and maximize the number of streams that your server can process, it is important to only run the EyesOnIt model when there might be something worth detecting.
EyesOnIt has three key features to help determine whether there might be something in your video stream worth analyzing with the EyesOnIt model. These features are: