Fast and Accurate Object Detection Based on Fusion of YOLOv2 and R-CNN Predicted Result for Autonomous Driving

Abstract

How to improve the safety of autonomous driving matters. Object detection could solve some problems at the perception level of autonomous driving. However, current models of object detection face the problem of low detection precision. In this paper, we propose a combined model, the fusion of You Only Look Once version 2(YOLOv2) and Regions With Convolutional Neural Networks(R-CNN) predicted results. Specifically, Regions With Convolutional Neural Networks(R-CNN) is used to predict positional parameters of a bounding box and You Only Look Once version 2(YOLOv2) is used to predict the class name and probability of a bounding box. The obtained results show that the combined model possesses higher precision than other compared models especially You Only Look Once version 2(YOLOv2) or Regions With Convolutional Neural Networks(R-CNN) individually.