People Counting in Public Spaces using Deep Learning-based Object Detection and Tracking Techniques

Recent advancements in deep learning and machine learning have enabled exact people counting in various applications including crowd management, security, and retail analytics. Deep learning algorithms have proven tremendous promise for accurate and efficient people counting in difficult contexts. This paper offers a technique for counting people that utilises deep learning with MobileNet SSD, centroid tracking, and trackable object script. Gathering and preparing a labelled dataset, training a MobileNet SSD model, implementing centroid tracking and the trackable object script, increasing system performance, testing it on real-world scenarios, and deploying it in a production environment are all part of the approach. The recommended approach provides a wide framework for creating and deploying a deep learning-based people-counting system that can be customized and tuned to match a number of applications and purposes. Additionally, we have added alerts on maximum capacity, timely scheduling and input feed from the internet.