Smart City Gnosys

Smart city article details

Title	Road Damage Detection Across Borders: Federated Learning Insights From Japan, China, Norway And The Usa
ID_Doc	46726
Authors	Dwivedi S.K.; Arya D.; Sekimoto Y.
Year	2024
Published	Proceedings - 2024 IEEE Smart World Congress, SWC 2024 - 2024 IEEE Ubiquitous Intelligence and Computing, Autonomous and Trusted Computing, Digital Twin, Metaverse, Privacy Computing and Data Security, Scalable Computing and Communications
DOI	http://dx.doi.org/10.1109/SWC62898.2024.00223
Abstract	This study explores the application of federated learning (FL) techniques to road damage detection across diverse geographical contexts, including Japan, Norway, China, and the USA. Federated learning, a decentralized machine learning approach, facilitates collaborative model training without sharing raw data, sharing only model parameters, thereby preserving privacy and enabling data synergy. While previous studies have been conducted on a smaller scale, using nearly similar datasets with consistent image capture methods, resolutions, and road views, our proposed work presents an extensive study of 4 countries (Japan, China, Norway, and the USA) demonstrating FL's effectiveness on a broader scale with diverse and heterogenous data addressing variations in road conditions, infrastructure, inspection methods, and environmental factors. The insights from the study reveal that the collaboratively trained model using FL outperforms the individual country's centralized models by a large margin. Leveraging YOLOv81 and the Flower framework with FedAvg strategy, our federated model garnered an mAP50 of 0.49 on a multi-country test dataset curated from these four participating countries, surpassing Japan's centralized model by 2.4% and outperforming models from Norway, China, and the USA approximately by about 20% and more. © 2024 IEEE.
Author Keywords	Automation; Big Data; Computer Vision; Deep Learning; Federated Learning (FL); Global Road Damage Detection; Intelligent Transport; Smart City Applications

Similar Articles

Id	Similarity	Authors	Title	Published
23667	0.974	Dwivedi S.K.; Arya D.; Sekimoto Y.	Enhanced Road Damage Detection With Federated Learning Across Diverse And Heterogeneous Global Datasets	32nd ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, ACM SIGSPATIAL 2024 (2024)
28082	0.878	Arya D.; Maeda H.; Kumar Ghosh S.; Toshniwal D.; Omata H.; Kashiyama T.; Sekimoto Y.	Global Road Damage Detection: State-Of-The-Art Solutions	Proceedings - 2020 IEEE International Conference on Big Data, Big Data 2020 (2020)
20646	0.871	Thamizharasi M.; Sethuraman R.; Sandhya A.	Distributed Framework For Pothole Detection And Monitoring Using Federated Learning: A Privacy-Preserving Edge Computing Approach	National Academy Science Letters (2025)
20645	0.87	Ezil Sam Leni A.; Shalen S.	Distributed Framework For Pothole Detection And Monitoring Using Federated Learning	2nd IEEE International Conference on Advances in Information Technology, ICAIT 2024 - Proceedings (2024)
16720	0.863	Arya D.; Maeda H.; Ghosh S.K.; Toshniwal D.; Omata H.; Kashiyama T.; Sekimoto Y.	Crowdsensing-Based Road Damage Detection Challenge (Crddc'2022)	Proceedings - 2022 IEEE International Conference on Big Data, Big Data 2022 (2022)
58038	0.862	Mehta S.; Singh A.	Towards Autonomous Road Maintenance: Cnn-Fl Integration For Pothole Detection Systems	2024 4th Asian Conference on Innovation in Technology, ASIANCON 2024 (2024)
204	0.861	Alshammari S.; Song S.	3Pod: Federated Learning-Based 3 Dimensional Pothole Detection For Smart Transportation	ISC2 2022 - 8th IEEE International Smart Cities Conference (2022)
44435	0.856	Kulambayev B.; Gleb B.; Katayev N.; Menglibay I.; Momynkulov Z.	Real-Time Road Damage Detection System On Deep Learning Based Image Analysis	International Journal of Advanced Computer Science and Applications, 15, 9 (2024)
40982	0.855	Arya D.; Omata H.; Maeda H.; Sekimoto Y.	Orddc'2024: State Of The Art Solutions For Optimized Road Damage Detection	Proceedings - 2024 IEEE International Conference on Big Data, BigData 2024 (2024)
31709	0.854	Awan A.Z.; Ji J.C.; Uzair M.; Ullah I.; Riaz W.; Gong T.	Innovative Road Distress Detection (Ir-Dd): An Efficient And Scalable Deep Learning Approach	PeerJ Computer Science, 10 (2024)