V-RoAst A New Dataset for Visual Road Assessment

This is dataset used for V-RoAst A New Dataset for Visual Road Assessment (https://arxiv.org/html/2408.10872v2)

The dataset used in this work is sourced from the ThaiRAP and was selected due to the availability of road assessment datasets adhering to the iRAP standard. The experiment includes 2,037 images (1600x1200 pixels) captured across Bangkok, Pathum Thani, and Phranakorn Sri Ayutthaya provinces, as illustrated in Figure 1 (please refer to the link). These images represent 519 road segments, with 1-4 images used to code each 100-meter segment. While iRAP datasets are typically not publicly available, this work will provide access to the ThaiRAP dataset, including both the images and their associated attributes, to support further research and development in road safety assessment.

Each segment has 52 attributes classified by iRAP. However, the attributes have a different number of classes (codes) in (parenthesis), and the number of classes of each attribute used in this experiment is shown at the top of each bar in Figure 2 (please refer to the link). This dataset shows the imbalance of classes of attributes, in which there are 11 attributes with only one class. And some classes have less than 10 samples.

To cite this dataset

@misc{jongwiriyanurak2024vroastnewdatasetvisual,
title={V-RoAst: A New Dataset for Visual Road Assessment},
author={Natchapon Jongwiriyanurak and Zichao Zeng and June Moh Goo and Xinglei Wang and Ilya Ilyankou and Kerkritt Srirrongvikrai and Meihui Wang and James Haworth},
year={2024},
eprint={2408.10872},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2408.10872},
}