AI-Based Analytics for Hawkers Identification in Video Surveillance for Smart Community

Abstract

Street hawking is a widespread phenomenon in urban areas globally, presenting challenges for local authorities such as traffic congestion, waste management, and negative impacts on the city's image. This research addresses key issues faced by authorities in managing hawkers, including the resistance to formalization, maintaining urban aesthetics, waste disposal, and understanding user preferences. The study investigates the performance of the You Only Look Once (YOLO) algorithm, utilizing Convolutional Neural Networks (CNN) for real-time object detection. To achieve thisobjective, the YOLOv5 algorithm is trained with a custom image dataset collected from the same camera along the street in the city area to detect five classes of objects, namely umbrella, table, stool, car, and people. Real images that were captured via camera and video surveillance were compiled as datasets which are then used to train and test the algorithm. The study aims to provide insights into the data collection process of hawkers along the street around the areas and the development of real-time hawker detection for the smart city application.