ABSTRACT
Crowd analysis is a new field of study that involves processing a large group of people to examine one or more of their behaviors. Deep learning is an appropriate technique for crowd analysis using a convolutional neural network. To calculate the distance between crowd members and to identify social distance violations, a deep crowd analysis is proposed in this study. Pre-trained in a single class To discover the region of interest, CNN is utilised to classify people (RoI). The people in the picture are then localized using a density map. The reference point used to calculate the distance between the people is the centroid of the isolated areas in the density map. A social distance violation is reported if the estimated distance is less than the specified threshold distance (3 meters). Between the two ROIs, a distance measured in pixels is determined. © 2022 IEEE.