Putra, Gilang Arya (2025) DETEKSI PELANGGARAN KENDARAAN MELAWAN ARAH MENGGUNAKAN ALGORITMA YOLO. S1 thesis, Universitas Jambi.
![]() |
Text
COVER.pdf Download (113kB) |
![]() |
Text
HALAMAN PENGESAHAN.pdf Download (330kB) |
![]() |
Text
ABSTRAK.pdf Download (132kB) |
![]() |
Text
BAB I.pdf Download (182kB) |
![]() |
Text
BAB V.pdf Download (122kB) |
![]() |
Text
DAFTAR PUSTAKA.pdf Download (160kB) |
![]() |
Text
SKRIPSI_FULL_GILANG ARYA PUTRA_F1E121210.pdf Restricted to Repository staff only Download (4MB) | Request a copy |
Abstract
Penelitian ini membahas tentang deteksi pelanggaran lalu lintas, khususnya pelanggaran kendaraan melawan arus, dengan memanfaatkan algoritma You Only Look Once (YOLO) sebagai metode pendeteksi objek secara real-time. Pelanggaran melawan arus merupakan salah satu faktor yang berkontribusi pada meningkatnya risiko kecelakaan lalu lintas, sehingga diperlukan upaya pemantauan yang efektif untuk menekan angka pelanggaran tersebut. Melalui pemanfaatan teknologi vision berbasis YOLO, penelitian ini mampu mengidentifikasi dan mengklasifikasikan kendaraan yang melakukan pelanggaran arah secara otomatis menggunakan dataset yang dikumpulkan di lingkungan Universitas Jambi. Proses penelitian meliputi tahap pengumpulan dan anotasi dataset, augmentasi gambar, pelatihan model dengan pengaturan hyperparameter yang sesuai, hingga evaluasi model menggunakan metrik mAP dan confusion matrix untuk mengukur performa deteksi. Dataset yang digunakan berjumlah 300 tiap kelas, terdapat 4 kelas dalam dataset yang digunakan berupa: mobil_depan, mobil_belakang, motor_depan dan motor_belakang. Akurasi terbaik diperoleh menggunakan hyperparameter berupa: epoch = 100, batch = 16, optimizer = AdamW, dropout = 0.0, learning rate(lr) = 0,001, IoU = 0.6, patience = 10, Cos_lr = True. Hasil penelitian menunjukkan bahwa model YOLOv11 yang telah dilatih mampu mendeteksi kendaraan yang melawan arus dengan akurasi optimal dan stabil pada pengujian data nyata, dengan nilai mAP yang tinggi dengan nilai 0.719 dan indikasi performa deteksi yang baik. This study discusses the detection of traffic violations, especially against-flow vehicle violations, by utilizing the You Only Look Once (YOLO) algorithm as a real-time object detection method. Against-flow violations are one of the factors that contribute to the increasing risk of traffic accidents, so effective monitoring efforts are needed to reduce the number of violations. Through the utilization of YOLO-based vision technology, this study is able to identify and classify vehicles that commit direction violations automatically using datasets collected at the University of Jambi. The research process includes the stages of dataset collection and annotation, image augmentation, model training with appropriate hyperparameter settings, to model evaluation using mAP metrics and confusion matrix to measure detection performance. The dataset used is 300 for each class, there are 4 classes in the dataset used, namely: front_car, rear_car, front_motorcycle and rear_motorcycle. The best accuracy was obtained using hyperparameters: epoch = 100, batch = 16, optimizer = AdamW, dropout = 0.0, learning rate(lr) = 0.001, IoU = 0.6, patience = 10, Cos_lr = True. The results showed that the trained YOLOv11 model was able to detect vehicles going against the flow with optimal and stable accuracy in real data testing, with a high mAP value of 0.719 and an indication of good detection performance.
Type: | Thesis (S1) |
---|---|
Uncontrolled Keywords: | Deep Learning, Object Detection, YOLO Algorithm, Traffic Violation, Real-Time Monitoring |
Subjects: | Q Science > Q Science (General) T Technology > T Technology (General) |
Divisions: | Fakultas Sains dan Teknologi > Sistem Informasi |
Depositing User: | GILANG ARYA PUTRA |
Date Deposited: | 11 Jul 2025 04:10 |
Last Modified: | 11 Jul 2025 04:10 |
URI: | https://repository.unja.ac.id/id/eprint/83622 |
Actions (login required)
![]() |
View Item |