Sensing, perception and decision for deep learning based autonomous driving

Abstract

Toward the realization of autonomous driving, deep learning has attracted the most attention, and it is seen as indispensable technology. AlexNet which consists of eight layers and incorporating ideas for improving generalization, was able to accomplish substantial improvement in accuracy with image recognition task. Since then, not only image recognition but also various tasks and applications to various fields are dramatically advanced. Even in the autonomous driving field, many manufacturers have taken aggressive efforts and are pushing ahead with practical application. In this paper, we will introduce the tasks being tackled for autonomous driving. The tasks introduced here are object detection, human pose estimation, and semantic segmentation from images and other sensors. By combining these methods, it is possible to realize safer automatic operation system.