An Improved Dilated Convolutional Network for Herd Counting in Crowded Scenes

TL;DR

This paper introduces an improved crowd counting model using a dual deep learning architecture with dilated convolutions and genetic algorithm optimization, achieving faster convergence and higher accuracy in congested scenes.

Contribution

The paper presents a novel two-part CNN system with optimized dilated rates for crowd counting, outperforming existing methods in speed and accuracy.

Findings

30% faster convergence than state-of-the-art

20% lower MAE on Shanghai dataset

Effective in highly congested scenes

Abstract

Crowd management technologies that leverage computer vision are widespread in contemporary times. There exists many security-related applications of these methods, including, but not limited to: following the flow of an array of people and monitoring large gatherings. In this paper, we propose an accurate monitoring system composed of two concatenated convolutional deep learning architectures. The first part called Front-end, is responsible for converting bi-dimensional signals and delivering high-level features. The second part, called the Back-end, is a dilated Convolutional Neural Network (CNN) used to replace pooling layers. It is responsible for enlarging the receptive field of the whole network and converting the descriptors provided by the first network to a saliency map that will be utilized to estimate the number of people in highly congested images. We also propose to utilize a genetic algorithm in order to find an optimized dilation rate configuration in the back-end. The proposed model is shown to converge 30\% faster than state-of-the-art approaches. It is also shown that it achieves 20\% lower Mean Absolute Error (MAE) when applied to the Shanghai data~set.