Vision-Language Models for Semantic Segmentation in Autonomous Driving : A Literature Review

Abstract

The objective of this thesis is to review the potential approaches for applying vision-language models to semantic segmentation within the context of autonomous driving. Rather than listing all suitable models, the focus is on exploring general implementation strategies. Each approach is examined in terms of its advantages and limitations, with relevant models used as examples to illustrate key points. The literature review indicates that while these methods show promise for autonomous driving applications, further development is needed before they can be effectively deployed in real-world scenarios. Overall, this work aims to provide a clearer understanding of the current landscape and identify potential areas for future research, contributing to the advancement of more reliable and practical perception systems for autonomous vehicles.