CRAFTSMAN

Algorithms using deep learning networks (or DNNs for Deep Neural Network) have recently attracted growing interest in both industry and academia.

Autonomous driving (or AD for Autonomous Driving) is one of the applications where DNN approaches have shown some level of performance.

However, to be effective in the case of AD, embedded systems for DNN must process a large amount of data from the various sensors in a limited amount of time and with minimum financial cost and energy consumption.

In this context, the ARTISAN project aims to develop new methods for exploring architectural optimizations to efficiently run a DNN application for DA. Our methods will make it possible to adapt the DNN algorithms used on the one hand, and the hardware architectures on the other, to different AD scenarios.