Smart sensors provide object classification and state estimation (position, speed, etc.) in real time. In the 5GMED project, the introduction of Artificial Intelligence-based smart sensors on the network might suppose stringent requirements for the network edge, including the necessity to process real-time video feeds.
The video below shows a solution that is able to be deployed at the extreme edge of the network (on the vehicle). Based on Machine Learning (YOLO v4), this smart sensor is able to classify vehicles, pedestrians and other agents commonly found in highway scenarios and determine whether they represent a hazard to traffic.
This sensor could be the solution to automotive use-cases where ubiquitous technology is needed to enforce safety. The video shows an example of these use cases and a working demo of the proposed smart sensor.