Crowd Level Monitoring System Using the R-CNN Mask Model

A crowd level detection system is a system that identifies crowds based on the number of visitors. This system makes it easy to calculate visitors automatically, and can provide information on crowd levels. In this research, we overcome the problem of inefficient manual supervision at Beringharjo Market by implementing human object detection technology. Using the Mask R-CNN method, this research aims to automatically identify and count the number of market visitors, improve monitoring and analysis of crowd levels. Mask R-CNN allows identifying people with high accuracy, precisely measuring crowds, and adding ROI lines to make calculations easier. Comparison with SSD shows Mask R-CNN has slightly lower accuracy with an accuracy of 0.8333 with SSD getting an accuracy of 0.927, but better object understanding. Mask R-CNN has a lower frame rate of 2 fps on average with SSD getting an average of 25 fps compared to SSD. The results of this research have the potential to increase the efficiency of crowd monitoring and analysis at Beringharjo Market, assisting market management in making strategic decisions and better planning.