Identifikasi Lampu Lalu Lintas Menggunakan Model NanoDet Pada Data Video

Danuarta, Valldynsyah (2025) Identifikasi Lampu Lalu Lintas Menggunakan Model NanoDet Pada Data Video. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5002211078-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (11MB) | Request a copy

Abstract

Pengembangan infrastruktur transportasi, terutama sistem lalu lintas, terus mengalami kemajuan di Indonesia. Sistem identifikasi lampu lalu lintas yang cepat dan akurat menjadi salah satu komponen penting dalam pengelolaan lalu lintas secara efektif. Identifikasi lampu lalu lintas secara otomatis dapat membantu optimalisasi sistem transportasi pintar dan mendukung implementasi kendaraan otonom. Pada penelitian Tugas Akhir ini berfokus pada eksplorasi salah satu model berbasis arsitektur Convolutional Neural Network (CNN) yaitu NanoDet untuk mengidentifikasi lampu lalu lintas secara real-time berdasarkan data video. NanoDet merupakan model identifikasi objek ringan dan cepat, yang dirancang untuk aplikasi real-time. Penelitian ini terdiri dari tiga tahapan utama. Tahap pertama adalah pengumpulan data primer, serta proses anotasi data. Tahap kedua adalah pelatihan model NanoDet untuk memperoleh model terbaik yang mampu mengidentifikasi lampu lalu lintas dan didapatkan performa mean average precision (mAP) sebesar 72.34%. Tahap ketiga adalah uji coba performa model untuk mengevaluasi kinerja model. Dari pengujian model memperoleh performa terbaik pada skenario cuaca tidak hujan dengan pencahayaan normal dengan precision sebesar 96.53%, recall sebesar 65.64%, f1-score sebesar 73.16%, accuracy sebesar 93.03%, dan FPS sebesar 68.07.
===================================================================================================================================
The development of transportation infrastructure, especially traffic systems, continues to progress in Indonesia. A fast and accurate traffic light identification system is one of the important components in effective traffic management. Automatic traffic light identification can help optimize smart transportation systems and support the implementation of autonomous vehicles. In this Final Project research focuses on exploring one of the architecture-based models, namely NanoDet, to identify traffic lights in real time based on video data. NanoDet is a lightweight and fast object identification model, designed for real-time applications. This research consists of three main stages. The first stage is primary data collection, as well as the data annotation process. The second stage is training the NanoDet model to obtain the best model capable of identifying traffic lights and obtaining the performance of the mean average precision (mAP) of 72.34%. The third stage is model performance testing to evaluate model performance. From the model testing, the best performance was obtained in a non-rainy weather scenario with normal lighting with a precision of 96.53%, recall of 65.64%, f1-score of 73.16%, accuracy of 93.03%, and FPS of 68.07.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Identifikasi, Lampu lalu lintas, NanoDet, Identification, Traffic light, NanoDet
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.6 Computer programming. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Valldynsyah Danuarta
Date Deposited:	22 Jul 2025 03:40
Last Modified:	22 Jul 2025 03:40
URI:	http://repository.its.ac.id/id/eprint/120431

Actions (login required)

View Item