Bird Eye View (BEV) and Dashcam Dataset Collection Using Semantic Segmentation and Depth Sensors on CARLA Simulator.

Rahmani, Thalita Aika and Widjanarko, Karla Vania (2026) Bird Eye View (BEV) and Dashcam Dataset Collection Using Semantic Segmentation and Depth Sensors on CARLA Simulator. Project Report. [s.n.]. (Unpublished)

Text 5025231058_5025231123-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Pengembangan sistem persepsi visual pada kendaraan otonom sangat bergantung pada ketersediaan data yang komprehensif untuk melatih model computer vision. Kendala utama dalam riset ini adalah tingginya biaya dan kompleksitas logistik dalam pengumpulan data fisik, terutama untuk memperoleh ground truth yang presisi, seperti label segmentasi pixel-wise dan estimasi kedalaman (depth) yang akurat. Selain itu, pelabelan manual untuk data semantik pada dunia nyata membutuhkan sumber daya manusia yang besar dan rentan terhadap human error. Simulasi digital menjadi solusi yang efisien untuk mengatasi keterbatasan tersebut. Dibandingkan simulator balap konvensional atau gim komersial, CARLA Simulator dirancang khusus untuk pengembangan kendaraan otonom dengan kemampuan menghasilkan data sintetik yang dilengkapi label segmentasi semantik dan informasi kedalaman secara otomatis serta akurat. Penelitian ini dilaksanakan di Laboratorium Pemodelan dan Komputasi Terapan (PKT), Departemen Teknik Informatika, Institut Teknologi Sepuluh Nopember (ITS), yang menyediakan infrastruktur komputasi berkinerja tinggi untuk mendukung simulasi real-time CARLA. Penelitian berjudul Bird Eye View (BEV) and Dashcam Dataset Collection Using Semantic Segmentation and Depth Sensors on CARLA Simulator bertujuan mengakuisisi dataset multimoda yang menggabungkan citra RGB, peta segmentasi semantik, dan peta kedalaman dari dua perspektif, yaitu Bird Eye View (BEV) dan Dashcam. Implementasi dilakukan menggunakan bahasa pemrograman Python melalui Jupyter Notebook pada Remote PC laboratorium. Dataset yang dihasilkan diharapkan dapat dimanfaatkan sebagai data pendukung penelitian lanjutan di bidang kecerdasan buatan, computer vision, dan pengolahan citra digital.

Kata Kunci: CARLA, Semantic Segmentation, Depth Sensor, Bird Eye View (BEV), Simulasi.
================================================================================================================================
The development of visual perception systems for autonomous vehicles highly depends on the availability of comprehensive datasets for training computer vision models. A major challenge in this field is the high cost and logistical complexity of collecting real-world data, particularly in obtaining precise ground truth information such as pixel-wise semantic segmentation labels and accurate depth estimation. Furthermore, manual annotation of semantic data requires significant human effort and is prone to human error. Digital simulation provides an efficient solution to these limitations. Unlike conventional racing simulators or commercial games, the CARLA Simulator is specifically designed for autonomous driving research and is capable of generating synthetic data with automatically produced and highly accurate semantic segmentation labels and depth information. This research was conducted at the Applied Modeling and Computation Laboratory (PKT), Department of Informatics, Institut Teknologi Sepuluh Nopember (ITS), which provides high-performance computing infrastructure to support real-time CARLA simulations. The study, entitled Bird Eye View (BEV) and Dashcam Dataset Collection Using Semantic Segmentation and Depth Sensors on CARLA Simulator, aims to acquire a multimodal dataset consisting of RGB images, semantic segmentation maps, and depth maps from two different perspectives, namely Bird Eye View (BEV) and Dashcam. The implementation was carried out using the Python programming language through Jupyter Notebook on the laboratory's Remote PC. The resulting dataset is expected to support future research in artificial intelligence, computer vision, and digital image processing.

Keywords: CARLA, Semantic Segmentation, Depth Sensor, Bird Eye View (BEV), Simulation.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	CARLA, Semantic Segmentation, Depth Sensor, Bird Eye View (BEV), Simulasi.
Subjects:	T Technology > T Technology (General) > T385 Visualization--Technique T Technology > T Technology (General) > T57.5 Data Processing T Technology > T Technology (General) > T57.62 Simulation T Technology > T Technology (General) > T59.7 Human-machine systems.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Thalita Aika Rahmani
Date Deposited:	10 Jul 2026 07:09
Last Modified:	10 Jul 2026 07:09
URI:	http://repository.its.ac.id/id/eprint/134604

Actions (login required)

View Item