2020/08/17 by Soufien Hamrouni, Hamrouni, Soufien, Hakim Ghazzai +5
Computer Science · #Video Surveillance and Tracking Methods #Anomaly Detection Techniques and Applications #Video Analysis and Summarization
paper · pdf · doi:10.48550/arxiv.2008.07254
Crowd management technologies that leverage computer vision are widespread in\ncontemporary times. There exists many security-related applications of these\nmethods, including, but not limited to: following the flow of an array of\npeople and monitoring large gatherings. In this paper, we propose an accurate\nmonitoring system composed of two concatenated convolutional deep learning\narchitectures. The first part called Front-end, is responsible for converting\nbi-dimensional signals and delivering high-level features. The second part,\ncalled the Back-end, is a dilated Convolutional Neural Network (CNN) used to\nreplace pooling layers. It is responsible for enlarging the receptive field of\nthe whole network and converting the descriptors provided by the first network\nto a saliency map that will be utilized to estimate the number of people in\nhighly congested images. We also propose to utilize a genetic algorithm in\norder to find an optimized dilation rate configuration in the back-end. The\nproposed model is shown to converge 30 % faster than state-of-the-art\napproaches. It is also shown that it achieves 20 % lower Mean Absolute Error\n(MAE) when applied to the Shanghai data~set.\n