Leveraging Deep Learning Vision for Real- Time Monitoring and Counting of Online Motorcycle Taxi Traffic

The expansion of online motorcycle taxi services has rapidly changed the transport systems in cities, especially in developing countries with high traffic density. This paper proposes a new method using deep learning and computer vision to monitor and count motorcycle taxis on the internet in real-time. The proposed system uses state-of-art convolutional neural networks (CNNs) for object detection and tracking, which the YOLOv5 model is chosen based on the combination of the accuracy and speed. The efficiency of the proposed system was evaluated in a real environment at the main gate of Universitas Brawijaya with the detection accuracy of 91,4% and reasonable time to complete the processes suitable for real time application. The versatility of the system was also demonstrated in different lighting and weather conditions with slight reductions in performance during low light and rainy environments. These promising results show the potential of the proposed system for improving traffic management in cities and supporting urban infrastructure development for smarter cities.