Analysis of Crowd Features based on Deep Learning

Everyday, crowds assemble for temples, concerts, marathons, anniversaries, memorials, holidays, political events, demonstrations, and shopping malls. In recent years, the automated assessment of crowd counting, gender, and age from face pictures has gained much interest due to its wide range of applications. Estimating people’s ages and gender in a crowd is difficult, but it has several applications in surveillance, management, and planning. A new approach to gender re-colonization, age estimation, and crowd counting based on gender with the location of individuals in visual frames is proposed in this study. Extended Mask R-CNN has been instructed to deal with all four issues simultaneously. Because localization necessitates the annotation of high-quality frames, created the SGSITS institute dataset, which overcomes the constraints of previous datasets. In addition, the study proposes evaluation methods and a full comparison with other approaches. With the training dataset of SGSITS, the proposed technique has outperformed state-of-the-art algorithms on a large dataset with exact annotations.