ABSTRACT
Crowd count estimation is one of the challenging jobs in the later age, because of non-uniform scale variations in densely crowded scenes. Key Application of crowd counting systems is public safety and protection. Other applications include, retail sectors, video surveillances etc. In current years, Deep convolution neural network has been shown effective at estimating crowd. Here, we introduce a new deep Convolution Neural Network (CNN) for estimating the crowd. This CNN learns both group groupings just as thickness map estimation which empowers the layers in the system to learn important discriminative highlights which help with assessing profoundly esteemed thickness maps with lower count error. Experiments on ShangaiTech data-set which is available publicly, show that the proposed method is efficient by gives lower count error and good quality density maps.
