Research on Crowd Counting and Density Estimation Algorithms Based on Deep Learning

Abstract

Thanks to the rapid development of computer vision technology, methods based on deep learning have gradually replaced counting methods based on traditional machine learning, and substantial progress has been made in counting accuracy and real-time detection. Firstly, the research background and application fields of target counting are introduced. Secondly, according to the classification of model tasks, the deep learning hotspot models are classified into three categories, and the crowd density estimation algorithms based on multi-scale strategies, multi-stage models and attention mechanisms, and multi-feature fusion are introduced from different perspectives. An introduction to the three algorithm models. Finally, it summarizes the shortcomings of the current target counting model, and looks forward to the future research directions.