A Saliency Based Human Detection Framework for Infrared Thermal Images

Abstract In this paper, a novel saliency framework for crowd detection in infrared thermal images is proposed. In order to obtain the optimal classifier from a large amount of data, the process of training consists of the following four steps: