A carregar...
Publicação
Synthetic data generation for lane detection: validation and analysis
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 113.7 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Abstract
Autonomous driving systems rely absolutely on lane detection. Therefore, ensuring its reliability is crucial for road safety. This work proposes validating one of the leading lane detection models in the CULane benchmark with an alternative synthetic dataset, with full automated ground truth labeling, from Epic Games’ Unreal Engine 5—a dynamically enriched, photorealistic simulation environment. By providing a range of diverse and challenging conditions (circadian, climatic, and road types), we aim to analyze the algorithm’s robustness and, in parallel, collect reference indicators of the domain gap versus real-world datasets. Results reinforce the role of synthetic data in expanding test coverage and minimizing the imbalance of training datasets for safetycritical applications.
Descrição
Dissertation submitted in partial fulfillment of the requirements for the
degree of Master of Science in Informatics and Multimedia Engineering
Palavras-chave
Synthetic data Domain gap Lane detection Autonomous driving
Contexto Educativo
Citação
COSTA, Pedro Cláudio Amaro da – Synthetic data generation for lane detection: validation and analysis. Lisboa: Instituto Superior de Engenharia de Lisboa. 2024. Dissertação de Mestrado.