| Issue |
EPJ Web Conf.
Volume 323, 2025
22nd International Metrology Congress (CIM2025)
|
|
|---|---|---|
| Article Number | 05001 | |
| Number of page(s) | 6 | |
| Section | Machine Learning | |
| DOI | https://doi.org/10.1051/epjconf/202532305001 | |
| Published online | 07 April 2025 | |
https://doi.org/10.1051/epjconf/202532305001
Estimating the repeatability and reproducibility of an AI-embedded measuring device: Application to road markings
1 CORE Center by COLAS, Data Science Digital Road Inspection and Material, 4 rue Jean Mermoz, F-78114, Magny-Les-Hameaux
2 ETIS UMR 8051, Cergy Paris University, ENSEA, CNRS, F-95000, Cergy, France
* Corresponding author: terence.bordet@colas.com
Published online: 7 April 2025
The road markings inspection using on-board cameras is gaining popularity, particularly for road inspection, despite most research focusing on autonomous vehicles. Traditional inspection methods rely on costly equipment like RetroTek-D, resulting in infrequent inspections. Recent advancements in artificial intelligence offer a more cost-effective solution, enabling the use of standard cameras with AI models for continuous road marking detection. While metrics such as F1-score evaluate AI performance, they do not ensure effectiveness on new sites or account for uncertainty in AI applications. Once deployed, AI systems may not adapt to new data, necessitating retraining for any updates. This paper proposes a method to estimate the repeatability and reproducibility of AI-based measurement devices, drawing inspiration from ISO 5725 standards, to enhance their reliability in road marking detection.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.