Faster R-CNN:an Approach to Real-Time Object Detection

Abstract

The objective of the paper is to present an example on how to use the latest image processing algorithms to detect traffic indicators safely enough to be used while driving a car. The conclusion of the paper is that the Faster Regional based Convolutional Neural Network (Faster R-CNN) algorithm has qualities in terms of accuracy and speed that make it suitable to be used in such applications. Faster R-CNN is a result of merging Region Proposal Network (RPN) and Fast-RCNN algorithms into a single network. For increasing the video processing power, a Graphics Processing Unit (GPU) was employed for training and testing at a speed of 15 fps on a dataset containing 3000 images for 4 classes. The dataset is composed of images containing the three phases of a traffic light and the STOP indicator.