Edge AI-Based Crowd Counting Application for Public Transport Stops

Abstract

Recently, the evolution of artificial intelligence has caused the emergence of smart systems exhibiting intelligent behavior like the human brain. Specifically, as a class of artificial intelligence methods, computer vision empowered with deep learning has tremendous promise for the accurate detection of crowds in real-time. In addition, the edge artificial intelligence approach allows for the development and deployment of artificial intelligence methods outside of the cloud. This study introduces the deep learning-based computer vision implementation to monitor public transport stops. The main aim is to determine the count of passengers through edge computing. The experimental study is realized with the popular YOLO object detector model on the Maixduino board developed for edge-based artificial intelligence (AI) applications with the internet of things (IoT). The experiments' results show that the obtained accuracy of crowd counting was found to be satisfactory.