Privacy-Preserving Multi-Camera Vehicle Detection for Smart Cities Using Federated YOLOv5s with UA-DETRAC

DOI

10.2991/978-94-6463-823-3_110How to use a DOI?

Keywords

Federated learning; vehicle detection; YOLOv5s; smart surveillance; edge computing

Abstract

With the increasing development of smart cities, large amounts of video data from surveillance systems pose significant challenges in terms of real-time processing, privacy protection, and efficient multi-camera collaboration. In the context of intelligent transportation systems, achieving accurate vehicle detection while preserving privacy and minimizing communication overhead is of great importance. To address these challenges, this paper proposes a federated You Only Look Once (YOLO) v5s-based framework that enables collaborative training across distributed camera nodes without requiring raw data sharing. By leveraging federated learning (FL), the framework allows each edge device to train its model locally on private data and only share model updates with a central server, effectively reducing privacy risks. To further improve the efficiency of the system, gradient quantization is applied during model aggregation, resulting in a significant 43% reduction in communication costs. The proposed framework is evaluated using the University of Alabama at Birmingham - Detection and Tracking of Vehicles in Aerial Camer (aUA-DETRAC) dataset, where it achieves a mean average precision (mAP@0.5) of 51.40%, demonstrating its effectiveness in detecting vehicles in various real-world scenarios. Moreover, the framework supports real-time inference at 32 frames per second (FPS) on edge devices, ensuring responsiveness for traffic monitoring tasks. In conclusion, this study presents a scalable and practical solution for privacy-aware vehicle detection in smart city environments. By combining the strengths of FL and lightweight object detection models, it offers a promising approach to balancing privacy, detection accuracy, and system efficiency in urban surveillance systems.

Copyright

© 2025 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yuxin Yang
PY  - 2025
DA  - 2025/08/31
TI  - Privacy-Preserving Multi-Camera Vehicle Detection for Smart Cities Using Federated YOLOv5s with UA-DETRAC
BT  - Proceedings of the 2025 3rd International Conference on Image, Algorithms, and Artificial Intelligence (ICIAAI 2025)
PB  - Atlantis Press
SP  - 1149
EP  - 1156
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6463-823-3_110
DO  - 10.2991/978-94-6463-823-3_110
ID  - Yang2025
ER  -