Human Detection with Three Dimensional Parts

Abstract—In this paper a human detection system based on parts of human body is proposed. We show privileged performance on people detection in the PASCAL VOC 2009 challenge. We use a three-dimensional part of annotations of humans in two dimension images with three-dimensional joint information. It enables to operate a hierarchical search method for finding parts that are closely clustered with three-dimensional information as well as two dimension image pixel values. For training every part SVM classifier is used with linear kernel which can be run over the image in a multi scale scanning mode and histogram of oriented gradients (HOG) features. The algorithm develops human detection system based on parts of humans such as head and shoulder views, among others.

Keywords:

Human detection , SVM , HOG , three-dimensional partdetection