Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{186993,
        author = {A.Gouri and Shruthi Peddapalli and G.Krishna Kaushik and T L Mokshanjali and Chiranjeevi Nuthalapati},
        title = {FederatedEdgeVision: Edge-Centric Deep Learning for Real-Time Visual Analytics and Privacy Preservation},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {6},
        pages = {3857-3867},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=186993},
        abstract = {The proliferation of real-time visual analytics demands a shift from traditional centralized cloud processing to decentralized, edge-centric paradigms. This paper proposes FederatedEdgeVision (FEV), an integrated framework leveraging Federated Learning (FL) to achieve real-time, high-accuracy visual analysis while adhering to stringent privacy mandates. FEV addresses the core trilemma of Accuracy-Privacy-Latency (APL) through three key innovations: 1) An Edge Optimization Pipeline integrating aggressive model quantization and lightweight CNN architectures (YOLOv8-Lite) for sub-100ms inference latency ; 2) A novel Personalized Federated Prototype Alignment (PFPA) algorithm to mitigate cross- client embedding-based data heterogeneity ; and 3) A layered privacy scheme combining Secure Aggregation (SecAgg) and optimized Differential Privacy (DP- FedAGS). Experiments conducted on non-IID partitioned COCO datasets confirm that FEV achieves a significant reduction in latency and a superior accuracy-privacy trade-off, quantified empirically using the Epsilon* privacy metric, demonstrating scalability for smart city, healthcare, and industrial safety applications.},
        keywords = {},
        month = {November},
        }