Real Time Crowd Counting System Using Machine Learning

Abstract

Crowd counting is a critical task in public safety, event management, and urban planning. This paper presents a real-time crowd counting system leveraging machine learning to accurately estimate the number of people in a given scene. The proposed system employs a convolutional neural network (CNN)-based deep learning model, optimized for processing images and video streams to identify and count individuals in diverse environments. Key features of the system include real-time inference, robust performance in varying lighting and density conditions, and adaptability to different camera perspectives. The model is trained on a diverse dataset, encompassing crowded events, open spaces, and public gatherings, ensuring its versatility and reliability. Post-training, the system is deployed using lightweight architectures, allowing seamless integration with edge devices and IoT platforms.