Situation-dependent selection of classification models based on spiking neural networks (3622)

►Reduction of energy requirements for the inference of machine learning methods in Edge AI applications, e.g. in vehicles

Based on a situation recognition (e.g. road type from map, vehicle speed, wetness detection, etc.), an orchestrator decides which of the implemented, trained and released (spiking) neural networks should currently be used to perform object detection, for example. These networks differ in terms of their energy requirements and performance, so that in simple situations the maximum performance is selected in terms of energy consumption and in demanding situations (e.g. when VRU is detected).