Pyramidal channel features for pedestrian detector

Abstract

Most state-of-the-art pedestrian detectors diversify features to improve the detection performance. Among such features, the gradient histogram is the most informative. This work investigates how diverse orientation information influences the pedestrian detection performance. In this paper, we propose a pedestrian detection method which uses pyramidal aggregated channel features which are generated with a layer-by-layer assembly scheme. Experimental results on the INRIA dataset show that our method can reach state-of-the-art performance despite using only gradient features. © 2015 ACM.