ITSC 2024 Paper Abstract

Close

Paper ThBT10.4

Shimomura, Kota (Chubu University), Inoue, Koki (Elith inc.), Ohmori, Kazuaki (Elith Inc.), Ryuta, Shimogauchi (Elith Inc.), Mimura, Ryota (Honda R&D Co., Ltd.), Ishikawa, Atsuya (Honda R&D Co., Ltd.), Kawabuchi, Takayuki (Honda R&D Co., Ltd.)

How to Extend the Dataset to Account for Traffic Risk Considering the Surrounding Environment

Scheduled for presentation during the Regular Session "Multi-autonomous Vehicle Studies, Models, Techniques and Simulations I" (ThBT10), Thursday, September 26, 2024, 15:30−15:50, Salon 18

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 14, 2024

Keywords Multi-autonomous Vehicle Studies, Models, Techniques and Simulations, Driver Assistance Systems, Cooperative Techniques and Systems

Abstract

To achieve safe autonomous driving, it is essential not only to perceive the surrounding environment but also to understand for various traffic risks in driving scenes. Previous studies have been limited to explanations of dynamic traffic risks and a limited range of static traffic risks. In particular, static traffic risks are explained without the consideration of road structures. However, extending (re-annotating) a dataset restricted to a specific use to an arbitrary use consumes an enormous amount of human resources. Therefore, this study proposes a method for automatically expanding the dataset to consider and explain static traffic risks in driving scenes using captions. Our approach selects regions with high traffic risks efficiently, utilizing GIS data and the explainability of machine learning models. Furthermore, by extracting road structure information with a high contribution to traffic accident risk predictions from explainability of machine learning models, it enables consideration of static traffic risks caused by road structures. Captions explaining traffic risks are generated using vision-language models (VLM) and prompts that can consider static risks. Experiments using the dataset constructed by proposed method demonstrate its ability to generate captions that consider static traffic risks for regions with high traffic risks on GIS data.

 

 

All Content © PaperCept, Inc.


This site is protected by copyright and trademark laws under US and International law.
All rights reserved. © 2002-2024 PaperCept, Inc.
Page generated 2024-10-14  02:09:16 PST  Terms of use