Deep Learning based on CNN for Pedestrian Detection: An Overview and Analysis

Abstract

Pedestrian detection methods based on deep learning can automatically learn features in unsupervised or supervised fashions and are capable to learn qualified high level feature representations to detect pedestrian. Recently, deep learning methods chiefly algorithms based on DCNN (Deep Convolutional Neural Network) have made outstanding achievements on pedestrian detection. The aim of this paper is to review the state-of-the-art in pedestrian detection algorithms based on DCNN. Also, we emphasize contributions and challenges of DCNN in pedestrian detection from the recent researches. In fact, this paper first overviews a number of well-known training approaches based on DCNN and their recent development, and then briefly describes recent algorithms based on these approaches. Ultimately, the paper analyzes deep pedestrian detection algorithms from training approach, categorization, and deep model points of view, and proposes a novel training method for deep pedestrian detection based on the analysis.