Crowd Density Identification at Cross Section using machine learning and deep learning

The estimation of crowd density at cross sections is a significant problem in the crowd analysis. In this study, we provide a novel technique for determining crowd density using Multi-column Convolutional Neural Networks (M-CNN). The proposed approach uses the M-CNN architecture to predict crowd densities at cross sections and learn hierarchical features from input photographs. Moreover, for this analysis ShanghaiTech dataset is utilized as (80-20)% training and testing datasets. The robustness of the M-CNN model is evaluated on the basis of normal error, Mean average error (MAE), Mean squared error (MSE), and growth truth. The findings demonstrate that the proposed model is highly accurate, and performs well after a few initial iterations. Furthermore, it can be incorporated in public places for crowd estimation and recognition.