Ni Gusti Putu Ayu Harry Saptarini
Politeknik Negeri Bali

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Sistem Pendeteksi Lampu Lalu Lintas untuk Kendaraan Pintar Menggunakan YOLOv9 Berbasis PyTorch Achmad Dewa; Ni Gusti Putu Ayu Harry Saptarini; I Putu Astya Prayudha
METHOMIKA: Jurnal Manajemen Informatika & Komputerisasi Akuntansi Vol. 10 No. 2 (2026): METHOMIKA: Jurnal Manajemen Informatika & Komputersisasi Akuntansi
Publisher : Universitas Methodist Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.46880/jmika.Vol10No2.pp543-553

Abstract

The ability to accurately detect traffic lights in real time is a crucial component for the safety and navigation of autonomous vehicles. This study aims to develop and evaluate a traffic light detection system (red, yellow, green) using the YOLOv9 deep learning algorithm based on the PyTorch framework. This study employs a quantitative approach with an experimental method utilizing the S2TLD dataset as well as additional data covering various lighting and weather conditions. The model was evaluated using Precision, Recall, Mean Average Precision (mAP), and inference time metrics. Test results show that the YOLOv9 model is highly effective, achieving a 98.0% mAP@0.5 in standalone testing and 96.5% when tested directly on a dashboard system using a local dataset. Additionally, the system recorded 0% fatal cross-color classification errors; the system recorded 0% cross-color classification errors on the test dataset, demonstrating a high level of reliability to support smart vehicle systems. In terms of processing speed, the system achieves an average inference time of 41.1 ms per frame (±24 FPS) using GPU acceleration. In conclusion, the YOLOv9 architecture has proven to be accurate, safe, and robust for implementation as an autonomous navigation perception system.