Driving Scene Understanding Using Hybrid Deep Neural Network

Abstract

Currently, artificial intelligence is used in many fields, especially real-time object recognition in vision is very efficient and powerful than other vision processing methods, and it is used for robots and autonomous vehicles. However, a more meaningful judgment by artificial intelligence requires recognizing the position of the detected object or the batch among the objects beyond the object detection and recognizing the situation. In other words, we can interpret that a person would measure a situation after seeing an object. In this paper, we propose a system to combine real-time object detection and situation recognition. We represent object detection and situation recognition by networks (DNN) with different characteristics. We propose an innovative system that can efficiently combine two networks.