Akurasi Pengujian Model Hasil Training menggunakan YOLOv4 untuk Pengenalan Kendaraan di Jalan Raya

Ahmad Fali Oklilas; sukemi; Dinda Dwinta; Ghinadhia Shofi; Nanda Putri Mariza; Sri Arum Kinanti; Yulia Amanda Sari

doi:10.5281./6537/15.jupiter.2023.04

Authors

Ahmad Fali Oklilas jurusan sistem komputer fakultas ilmu komputer universitas sriwijaya
sukemi jurusan sistem komputer fakultas ilmu komputer universitas sriwijaya
Dinda Dwinta Jurusan Sistem Komputer Fakultas Ilmu Komputer Universitas Sriwijaya
Ghinadhia Shofi Jurusan Sistem Komputer Fakultas Ilmu Komputer Universitas Sriwijaya
Nanda Putri Mariza Jurusan Sistem Komputer Fakultas Ilmu Komputer Universitas Sriwijaya
Sri Arum Kinanti Jurusan Sistem Komputer Fakultas Ilmu Komputer Universitas Sriwijaya
Yulia Amanda Sari Jurusan Sistem Komputer Fakultas Ilmu Komputer Universitas Sriwijaya

DOI:

https://doi.org/10.5281./6537/15.jupiter.2023.04

Abstract

Traffic congestion is currently the main problem that occurs in big cities in Indonesia.
Traffic flow analysis is an important basis for urban planning. Management of Intelligent
Transportation System (ITS) has become a necessity today to manage heavy traffic problems.
Intelligent transportation systems using computer vision techniques are increasingly attracting
attention for traffic density detection. This research uses the You Only Look Once (YOLO version
4 object detection method for vehicle classification and detection to obtain an optimal model.
Testing the YOLOv4 model results in a mean average precision (mAP) of 80.12%. In video testing
to detect motorcycles and cars, the total vehicle accuracy is 70.6% and the vehicle confidence
level is 78.7%.