This paper explores the application of federated learning (FL) in road damage detection, addressing the challenges of data privacy and large-scale data transfer. The study uses the YOLOv5l object detection model and the Road Damage Detection Dataset (RDD2022) to train models for single-country and multi-country applications. FL allows multiple clients to collaboratively train models without sharing their data, enhancing privacy and reducing data transfer. The results show that FL models, while achieving 21%–25% lower mean average precision (mAP) compared to centralized models, outperform local client models by 1.33%–163% on global test data. The study also analyzes the impact of data distribution, number of clients, server rounds, and local epochs on model performance, highlighting the importance of careful parameter selection to balance communication costs and model accuracy. The findings suggest that FL can be a robust solution for road damage detection, particularly in scenarios where data privacy and diverse data sources are crucial.This paper explores the application of federated learning (FL) in road damage detection, addressing the challenges of data privacy and large-scale data transfer. The study uses the YOLOv5l object detection model and the Road Damage Detection Dataset (RDD2022) to train models for single-country and multi-country applications. FL allows multiple clients to collaboratively train models without sharing their data, enhancing privacy and reducing data transfer. The results show that FL models, while achieving 21%–25% lower mean average precision (mAP) compared to centralized models, outperform local client models by 1.33%–163% on global test data. The study also analyzes the impact of data distribution, number of clients, server rounds, and local epochs on model performance, highlighting the importance of careful parameter selection to balance communication costs and model accuracy. The findings suggest that FL can be a robust solution for road damage detection, particularly in scenarios where data privacy and diverse data sources are crucial.