Crowd counting analysis using deep learning: a critical review

Abstract

The term "crowd counting" refers to the practise of counting the number of people present in a certain area. Urban planning, medical services, emergency preparedness, public security, strategic planning, and defence all seem to be domains where this method may be used. Occlusion, size and perspective distortion, and non-uniform distribution are all problems that crowd counting approaches face. As the population density grows, so does the complexity of the calculations. Great advances in deep convolution neural networks (CNNs) and datasets are largely responsible for the tremendous development in crowd count approaches seen in the past few years. In this paper we assess recent efforts and provide a complete evaluation of modern deep learning-based crowd counting systems. This paper discusses some classic and deep learning-based crowd counting approaches. We examine detection-based, regression-based, and classic density estimation approaches briefly. For the purpose of estimating the crowd density and count for the provided crowd scene image, we have evaluated the recent 10 publications on crowd counting using deep learning. We also go through the most widely used datasets. In conclusion, these investigations demonstrated a high degree of precision and offered good illustrations of the potential of AI in crowd counting. From paper to paper, there were significant differences in the approaches and algorithms used to address the crowd counting and density mapping challenge. We also examine the possible uses of crowd counting as well as the difficulties associated with it. As a result, fresh study is being conducted in this area.