Performance Evaluation of Ambulance Detection and Classification Using Machine Learning Through a Drone on Edge AI Devices

Emergency services like ambulances, fire brigades, etc. are essential pillars of any country. In certain situations, the delayed arrival of ambulances at hospitals can pose a risk to the health and well-being of patients. This delay often arises from traffic congestion. Nowadays, with the rise in traffic, it is essential to rapidly spot and recognize ambulances to facilitate emergency services more quickly and save lives. In this paper, we have proposed an approach to detect and identify ambulances using machine learning techniques. We employ various versions of You Only Look Once (YOLO) object detection algorithms to detect ambulances from drone visions and deploy the machine learning models on two distinct Edge AI hardware platforms: Intel's UP Squared board and Nvidia Jetson Nano for real-time inference. The models are evaluated using a dedicated test dataset and also tested for the detection of an ambulance through drone videos in real-time. The experimental results successfully demonstrate the effectiveness of our approach in real-world scenarios. The mean Average Precision (mAP50-90) values obtained are 0.723 for YOLOv5 and 0.979 for YOLOv8, indicating their respective performance for Ambulance detection using drones and real-time inference. This can be beneficial in practical applications including smart emergency services, traffic management, and transportation systems. Our research contributes to the field of ambulance detection and classification, offering insights into the effectiveness of machine-learning techniques.