Kumar, Nikhil and Lal, Krishan and Singh, Geeta and Sharma, Anuraganand (2024) Fijian Traffic Sign Dataset: A New Collection for Image Recognition and Benchmarking. [Conference Proceedings] (In Press)
Full text not available from this repository. (Request a copy)Abstract
This research work analyses the performance of a Convolutional Neural Network (CNN) based traffic sign recognition model using a dataset of Fijian traffic signs collected under various environmental conditions. The aim is to contribute to the increasing need for not only reliable and accurate road sign recognition systems but also safe systems for human drivers as well as autonomous systems. Our dataset consists of six different types of Fiji’s traffic signs and incorporates various environmental conditions such as quality, fog, and motion blur to mimic real-world scenarios. Using the Fijian dataset ensures that the recognition system is tailored to the specific characteristics of Fijian road signs and local conditions, enhancing regional applicability and compliance with local traffic regulations. The experimental results reveal that adverse conditions significantly impact the model’s accuracy. For example, the classification accuracy ranges from 86.80%-96.44%, 73.53%-96.30%, and 84.78%-96.55% for image quality, fog intensity and speed impact. This study also highlights the limitations of current traffic sign recognition systems and offers insights into future enhancements to increase their resilience and dependability in real-world applications.
| Item Type: | Conference Proceedings |
|---|---|
| Additional Information: | Program schedule: https://iconip2024.org/programme/ |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science |
| Divisions: | Faculty of Science, Technology and Environment (FSTE) > School of Computing, Information and Mathematical Sciences |
| Depositing User: | Anuraganand Sharma |
| Date Deposited: | 20 Jul 2025 22:33 |
| Last Modified: | 20 Jul 2025 22:33 |
| URI: | https://repository.usp.ac.fj/id/eprint/14801 |
Actions (login required)
 |
View Item |