Domain generalized federated learning for Person Re-identification
Lists
Only you see this
Abstract
Highlights•Unified privacy and generalization. We propose a novel framework that seamlessly integrates decentralized learning and a Domain Compensation Network (DCN) to jointly address privacy concerns and enhance domain generalization in ReID tasks.•Synergistic DCN components. We introduce a Domain Compensation Network (DCN) within each client, comprising domain generation, diversity regularization, and fidelity regularization, aiming to simulate unseen domains, capture diverse clues, and preserve discriminability for robust generalization.•Holistic decentralized learning. We Utilize local updating and dynamic aggregation strategies within decentralized federated learning, ensuring privacy preservation and robust domain generalization.
BibTeX
@article{liu2024domain,
author = {Fangyi Liu and Mang Ye and Bo Du},
title = {{Domain generalized federated learning for Person Re-identification}},
journal = {Comput. Vis. Image Underst.},
year = {2024},
doi = {10.1016/j.cviu.2024.103969},
}Loading related content...