Crowd counting using complex convolutional neural network
Marcin Matlacz, Grzegorz Sarwas · 2018
This paper is focused on the problem of counting people in crowd. For solving this issue a complex valued convolutional neural network has been proposed. The network training and evaluation have been processed using datasets ShanghaiTech and UCF_CC_50, respectively. Achieved results have been compared with other algorithms for crowd counting based on the deep neural network architecture, mainly “CrowdNet” algorithm. Proposed model achieved better results than equivalent real-valued model.