Putra, Alfadli Rahmat (2025) Deteksi Penggunaan Helm pada Pengendara Sepeda Motor dengan YOLO11. S1 thesis, Universitas Jambi.
![[img]](https://repository.unja.ac.id/style/images/fileicons/text.png) |
Text
COVER.pdf Download (74kB) |
|
Text
HALAMAN PENGESAHAN.pdf Download (431kB) |
|
Text
ABSTRAK.pdf Download (87kB) |
|
Text
BAB I.pdf Download (123kB) |
|
Text
BAB V.pdf Download (90kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (139kB) |
|
Text
FULL SKRIPSI.pdf Restricted to Repository staff only Download (10MB) |
Abstract
Kecelakaan lalu lintas merupakan permasalahan serius yang terus meningkat, baik secara global maupun di Indonesia, dengan sepeda motor sebagai penyumbang utama. Penggunaan helm sebagai alat pelindung diwajibkan dalam peraturan perundang-undangan untuk mengurangi risiko cedera, namun pengawasan manual terhadap kepatuhannya masih belum optimal. Oleh karena itu, penelitian ini mengembangkan sistem pendeteksi pelanggaran helm berbasis teknologi deteksi objek YOLO (You Only Look Once). Dataset dikumpulkan dari CCTV ATCS Kota Samarinda dalam bentuk video dan gambar, kemudian diproses menjadi frame statis, dibersihkan, dianotasi (helmet, no helmet, rider, motorcycle), di-resize, dibagi, dan diaugmentasi. Model dilatih dengan pendekatan transfer learning pada delapan konfigurasi yang divariasikan berdasarkan epoch, freeze, dan optimizer. Hasil evaluasi menunjukkan performa terbaik diperoleh dari model dengan konfigurasi SGD, tanpa freeze, dan 50 epoch, menghasilkan precision 0,950, recall 0,971, mAP50 sebesar 0,972, dan mAP50-95 sebesar 0,753. Model ini diimplementasikan ke dalam antarmuka berbasis Streamlit untuk pemantauan pelanggaran helm secara otomatis dan real-time. Traffic accidents remain a serious and growing issue worldwide and in Indonesia, with motorcycles being a major contributor. Although helmet use is legally mandated to reduce injury risks, manual monitoring of compliance is still suboptimal. This study develops a helmet violation detection system using YOLO (You Only Look Once) object detection technology. The dataset was collected from ATCS CCTV footage in Samarinda, consisting of images and videos, which were converted into static frames, cleaned, annotated (helmet, no helmet, rider, motorcycle), resized, split, and augmented. The model was trained using a transfer learning approach with eight different configurations varying in epoch count, freeze status, and optimizer type. Evaluation results showed the best performance using the SGD optimizer, with no freeze and 50 epochs, achieving a precision of 0.950, recall of 0.971, mAP50 of 0.972, and mAP50-95 of 0.753. The final model was deployed into a Streamlit-based interface to enable practical and real-time helmet violation monitoring.
| Type: | Thesis (S1) |
|---|---|
| Uncontrolled Keywords: | Helmet Violation, YOLO, Object Detection, Transfer Learning, Real-time Monitoring |
| Subjects: | Q Science > Q Science (General) T Technology > T Technology (General) |
| Divisions: | Fakultas Sains dan Teknologi > Sistem Informasi |
| Depositing User: | ALFADLI RAHMAT PUTRA |
| Date Deposited: | 11 Jul 2025 04:13 |
| Last Modified: | 11 Jul 2025 04:13 |
| URI: | https://repository.unja.ac.id/id/eprint/83381 |
Actions (login required)
 |
View Item |
