<![CDATA[Abstrak − Penelitian ini mengembangkan sistem deteksi otomatis pelanggaran lalu lintas pengendara sepeda motor metik dan bebek menggunakan You Only Look Once (YOLOv11n). Sistem menargetkan pelanggaran seperti tanpa helm, tanpa plat nomor, tanpa spion, dan berbonceng lebih dari dua orang. Dataset sebanyak 739 gambar, diaugmentasi melalui rotasi dan flipping, dilatih hingga 500 epoch. Performa terbaik dicapai pada epoch 350 dengan mAP@0.5 0,572 dan mAP@0.5:0.95 0,397. Ketidakseimbangan data memengaruhi deteksi kelas minor. Sistem ini menjanjikan untuk pengawasan lalu lintas otomatis, dengan rekomendasi untuk meningkatkan kualitas data dan eksplorasi YOLO varian terbaru.Kata Kunci: YOLOv11n; pelanggaran lalu lintas; deteksi objek; sepeda motor; mAP;Abstract — This study develops an automatic traffic violation detection system for riders of moped and underbone motorcycles using You Only Look Once (YOLOv11n). The system targets violations such as riding without a helmet, without a license plate, without mirrors, and carrying more than two passengers. A dataset of 739 images was augmented through rotation and flipping and trained for up to 500 epochs. The best performance was achieved at epoch 350 with mAP@0.5 of 0.572 and mAP@0.5:0.95 of 0.397. Data imbalance affected the detection of minor classes. This system shows promise for automatic traffic surveillance, with recommendations to improve data quality and further explore newer YOLO variants.Keywords: YOLOv11n; traffic violation; object detection; motorcycle; mAP;]]>
Copyrights © 2025
