Image Classification Based on Federated Learning and PFLlib

Weiqing Fu

2024

Abstract

Image classification has always been a research hotspot in the computer vision community, which is also the foundation of many higher-order scene understanding tasks. Based on data-driven ideas, most existing image classification models rely on massive data and centralized large-scale training. However, due to the security and privacy issues of data, practical application scenarios often cannot fully utilize all training data, resulting in significant room for improvement in the accuracy and robustness of the model. Inspired by the rapid development of federated learning, this article introduces the idea of local training and global updates into image classification tasks, exploring the performance boundaries of different representative federated learning algorithms in classification tasks. Specifically, based on the PFLlib platform, this article designs a unified Client and Server end that can integrate common federated learning algorithms. In addition, this article quantitatively compares the impact of neural network structures on the classification performance of different methods. Extensive experiment results have verified the significant improvement of federated learning in classification performance.

Paper Citation

in Harvard Style

Fu W. (2024). Image Classification Based on Federated Learning and PFLlib. In Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI; ISBN 978-989-758-713-9, SciTePress, pages 646-651. DOI: 10.5220/0012961200004508

in Bibtex Style

@conference{emiti24,
author={Weiqing Fu},
title={Image Classification Based on Federated Learning and PFLlib},
booktitle={Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI},
year={2024},
pages={646-651},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012961200004508},
isbn={978-989-758-713-9},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence - Volume 1: EMITI
TI - Image Classification Based on Federated Learning and PFLlib
SN - 978-989-758-713-9
AU - Fu W.
PY - 2024
SP - 646
EP - 651
DO - 10.5220/0012961200004508
PB - SciTePress