Information Entropy Augmented High Density Crowd Counting Network

The research proposes an innovated structure of the density map-based crowd counting network augmented by information entropy. comprises a front-end to extract features and back-end generate maps. In order validate assumption that entropy can boost accuracy map generation, multi-scale extraction process is imported into along with fine-tuned convolutional feature process, network, extracted are decoded multi-column dilated convolution network. Finally, be mapped as estimated number. Experimental results indicate devised capable accurately estimating count in extremely high density. Compared similar structured networks which don’t adapt feature, proposed exhibits higher performance. This result proves enhancing efficiency approaches.