Sistem Deteksi Pelanggaran Kendaraan Mobil dan Sepeda Motor di Zona Larangan Parkir menggunakan YOLOv8 Berbasis Web

Illegal parking in no-parking zones often causes traffic and road transport disruptions, thus requiring a technology-based solution to automatically detect such violations. This study aims to develop a parking violation detection system for two-wheeled and four-wheeled vehicles using the YOLOv8 object detection model. The system development employs the prototype method, which allows for iterative creation and testing to achieve optimal results. The system is implemented on IoT devices (Raspberry Pi 4, webcam, buzzer) and integrated with a Laravel-based web dashboard for monitoring violations. The YOLOv8 model is trained on a dataset and evaluated using precision, recall, and mean Average Precision (mAP) metrics at Intersection over Union (IoU) thresholds of 50% (mAP50) and 50–95% (mAP50-95), as well as inference speed to assess real-time capability. Evaluation results show the model achieves a mAP50 of 96.3% with high precision, although recall for the motorcycle class is lower compared to the car class. The system is capable of providing real-time alerts via buzzer when a parking violation is detected, and displaying violation data on the web dashboard. The YOLOv8-based parking violation detection system has been successfully implemented through prototype development, and the system operates well according to expectations and predefined specifications.
Keyword : Parking Violation, Deep Learning, YOLOv8, Internet of Things, Laravel