Using Channel Feature with RPN and SVM for Pedestrian Detection

Abstract

Detecting pedestrian has been arguably addressed, especially in the fields of automotive, surveillance and robotics. Despite the significant improvements, pedestrian detection is still an open challenge that calls for more and more accurate algorithms. In this paper, we propose a pedestrian detection system based on deep learning, adapting Region Proposal Network (RPN) of the faster CNN to generate the region proposals, and then utilizing the VGG16 convolutional layer to obtain the feature vector of the channel feature and normalize them, finally putting them into SVM classifier. By thoroughly analyzing and optimizing each step of the detection pipeline we propose an architecture that outperforms most of the traditional methods. Our approach can achieve 8.95% miss rate on Caltech pedestrian dataset and it is competitive.