Examining the Crowd in Real-time with Deep Learning

Abstract

In this research, a model is proposed that can be used to estimate crowd density in a specific region and to establish social distances in accordance with predetermined rules. This is accomplished utilizing a multi-source model-based approach. In a small public gathering where hand counting is impossible, this technique conducts a survey. To do this, input video frames are extracted, each frame is processed, and then passed to the model for additional detection and implementation. To find people in the provided survey feed, the idea of object detection is applied. The model then determines how many people are in the vicinity. The Euclidean metric is also used to calculate the distance between people who are present there. These criteria help to ensure that the social-distancing rules and procedures specific to the area are being followed. The suggested model is based on YOLO, a deep learning algorithm that detects objects (in this case, people) and aids in classifying the necessary parameters for this project. The study will go into great detail about this algorithm.