Real-Time ID Card Detection Using Different YOLO Models

A secure environment within educational institutions is a critical concern, especially given the diverse demographic that includes students, faculty, and staff. The capability to accurately identify individuals through identity cards is integral to ensuring safety, particularly during high-traffic periods such as admissions and large-scale events. This paper offers a comprehensive analysis of the YOLO V7, V8, and V9 models for real-time identity card detection, Utilizing cutting-edge deep learning techniques. The dataset used is 1037 images from Roboflow Universe, “dress-code-detection-myycl”, which includes images of people without and with ID cards, as well as with changes in clothes that can be classified by gender. The working methodology followed will involve dataset annotation, training three models of YOLO with given configurations of input size, epochs, loss functions, and image augmentation, and testing their performances based on precision, recall, mAP50, and mAP50:95. The results indicate that YOLOv8 achieved the highest mAP50:95 of 72.63%, outperforming YOLOv9 (70.69%) and YOLOv7 (38.42%). This suggests that YOLOv8 demonstrates superior accuracy in real-time ID card detection.The research evaluates the performance of each model in terms of accuracy, processing speed, and robustness, providing a detailed evaluation of their practical applications in a campus security context. By enhancing identification processes, these models aim to bolster security protocols, mitigate potential threats, and streamline daily operations. This study not only emphasizes the pivotal role of deep learning in educational security but also provides actionable insights into the optimal model for effective identity card detection, contributing to the advancement of safe and efficient educational environments.