An artificial intelligence (AI) system for automotive radar sensors developed by researchers at Austria's Graz University of Technology (TU Graz) filters out interfering signals from other radar sensors to improve object detection.
The researchers built model architectures for automatic noise suppression based on convolutional neural networks (CNNs).
To make them more efficient, the researchers trained the neural networks with noisy data and desired output values, then compressed the most efficient models further by reducing bit widths, resulting in an AI model with high filter performance and low energy consumption.
Said TU Graz's Franz Pernkopf, "We want to make CNNs’ behavior a bit more explainable. We are not only interested in the output result, but also in its range of variation. The smaller the variance, the more certain the network is.”
From Graz University of Technology (Austria)
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA