Penggunaan helm oleh pengendara sepeda motor memiliki peran vital dalam mengurangi risiko cedera fatal, khususnya di daerah padat pengendara bermotor dan pejalan kaki. Penelitian ini mengembangkan sistem deteksi otomatis penggunaan helm berbasis algoritma You Only Look Once versi 8 (YOLOv8). Data diperoleh dari hasil pengambilan gambar menggunakan kamera ponsel, kemudian dianotasi melalui Roboflow dengan label “helm” dan “tanpa helm”. Untuk meningkatkan kemampuan generalisasi model, dataset diproses menggunakan teknik augmentation, meliputi rotation, flipping, perubahan skala, dan penyesuaian warna. Proses pelatihan dan pengujian dilakukan di Google Colab dengan dukungan GPU, menggunakan pembagian data pelatihan dan validasi. Evaluasi performa dilakukan melalui metrik precision, recall, F1-score, dan mean Average Precision (mAP). Hasil pengujian menunjukkan bahwa YOLOv8 mampu mendeteksi penggunaan helm secara real-time dengan akurasi tinggi pada berbagai kondisi pencahayaan dan sudut pandang. Penelitian ini membuktikan bahwa YOLOv8 berpotensi menjadi solusi efektif untuk pemantauan keselamatan lalu lintas, serta dapat diintegrasikan ke dalam sistem pengawasan berbasis kamera atau aplikasi mobile untuk mendukung penegakan peraturan keselamatan berkendara.

The use of helmets by motorcycle riders plays a vital role in reducing the risk of fatal injuries, particularly in areas with high traffic density involving both motorcyclists and pedestrians. This study develops an automatic helmet detection system based on the You Only Look Once version 8 (YOLOv8) algorithm. The dataset was obtained from images captured using a mobile phone camera and annotated through Roboflow with the labels “helmet” and “no helmet.” To enhance the model’s generalization ability, the dataset was augmented using techniques such as rotation, flipping, scaling, and color adjustment. The training and testing processes were conducted in Google Colab with GPU support, using a split between training and validation data. Model performance was evaluated using precision, recall, F1-score, and mean Average Precision (mAP). The experimental results show that YOLOv8 can detect helmet usage in real-time with high accuracy under various lighting conditions and camera angles. This research demonstrates that YOLOv8 has strong potential as an effective solution for traffic safety monitoring and can be integrated into camera-based surveillance systems or mobile applications to support the enforcement of road safety regulations.