Detection and Classification of Emergency Vehicles using Deep Learning

Emergency vehicle detection plays a critical role in ensuring safe, efficient, and sustainable transportation within modern urban environments. Unfortunately, delays in their response times can have significant consequences. This project proposes a novel approach to address this challenge by leveraging state-of-the-art deep learning techniques for emergency vehicle detection within an Intelligent Transportation Management System (ITMS) framework. The proposed system utilizes a two-step approach. First, it employs the You Only Look Once (YOLO) algorithm, a real-time object detection system, to identify and localize vehicles captured by traffic surveillance cameras. This allows the system to isolate vehicles within the image data for further analysis. Following vehicle detection, the system performs classification using three deep learning architectures: MobileNet, EfficientNet, and ResNet. Each of these models will be evaluated to determine its effectiveness in accurately and efficiently classifying vehicle types, with a particular focus on identifying emergency vehicles. By comparing the performance of these models, the project will identify the optimal choice for real-world implementation within the ITMS. Finally, to further enhance the chosen classification model's performance, the project will utilize the Adam optimizer. This optimizer offers improved convergence behavior and robustness compared to traditional methods, leading to a more accurate and reliable system.