Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AdaFedFR: Federated Face Recognition with Adaptive Inter-Class Representation Learning

Di Qiu, Xinyang Lin, Kaiye Wang, Xiangxiang Chu, Pengfei Yan

TL;DR

This paper addresses privacy-aware federated face recognition under non-IID data and communication constraints. It introduces AdaFedFR, which leverages shareable mean feature embeddings of public identities as global class representations and an adaptive $k$-negative contrastive loss to align local and global representations, along with a local adapter and differential privacy. The method achieves faster convergence and improved generic and personalized performance, outperforming prior FL FR methods on major benchmarks within fewer communication rounds. This approach reduces communication and computation overhead while enhancing privacy protections, making federated FR more practical for real-world deployments.

Abstract

With the growing attention on data privacy and communication security in face recognition applications, federated learning has been introduced to learn a face recognition model with decentralized datasets in a privacy-preserving manner. However, existing works still face challenges such as unsatisfying performance and additional communication costs, limiting their applicability in real-world scenarios. In this paper, we propose a simple yet effective federated face recognition framework called AdaFedFR, by devising an adaptive inter-class representation learning algorithm to enhance the generalization of the generic face model and the efficiency of federated training under strict privacy-preservation. In particular, our work delicately utilizes feature representations of public identities as learnable negative knowledge to optimize the local objective within the feature space, which further encourages the local model to learn powerful representations and optimize personalized models for clients. Experimental results demonstrate that our method outperforms previous approaches on several prevalent face recognition benchmarks within less than 3 communication rounds, which shows communication-friendly and great efficiency.

AdaFedFR: Federated Face Recognition with Adaptive Inter-Class Representation Learning

TL;DR

This paper addresses privacy-aware federated face recognition under non-IID data and communication constraints. It introduces AdaFedFR, which leverages shareable mean feature embeddings of public identities as global class representations and an adaptive -negative contrastive loss to align local and global representations, along with a local adapter and differential privacy. The method achieves faster convergence and improved generic and personalized performance, outperforming prior FL FR methods on major benchmarks within fewer communication rounds. This approach reduces communication and computation overhead while enhancing privacy protections, making federated FR more practical for real-world deployments.

Abstract

With the growing attention on data privacy and communication security in face recognition applications, federated learning has been introduced to learn a face recognition model with decentralized datasets in a privacy-preserving manner. However, existing works still face challenges such as unsatisfying performance and additional communication costs, limiting their applicability in real-world scenarios. In this paper, we propose a simple yet effective federated face recognition framework called AdaFedFR, by devising an adaptive inter-class representation learning algorithm to enhance the generalization of the generic face model and the efficiency of federated training under strict privacy-preservation. In particular, our work delicately utilizes feature representations of public identities as learnable negative knowledge to optimize the local objective within the feature space, which further encourages the local model to learn powerful representations and optimize personalized models for clients. Experimental results demonstrate that our method outperforms previous approaches on several prevalent face recognition benchmarks within less than 3 communication rounds, which shows communication-friendly and great efficiency.
Paper Structure (23 sections, 8 equations, 2 figures, 6 tables, 1 algorithm)

This paper contains 23 sections, 8 equations, 2 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Face Recognition
  4. Federated Learning
  5. Federated Learning in Face Recognition
  6. Methodology
  7. Problem Setup
  8. Motivation
  9. Framework Overview
  10. Shareable Representations
  11. Representation Learning in Feature Space
  12. Local Objective
  13. Experiments
  14. Experimental Settings
  15. Method Comparison
  16. ...and 8 more sections

Figures (2)

  • Figure 1: Overview of our proposed federated face recognition framework. At each communication round, we optimize the local model with local data and global class representations by utilizing a combined objective consisting of cosface loss, adaptive k-negative contrastive loss and bce loss, which effectively enlarge the distances of different identities (i.e. $r_{c1}$ and $r_{s1}$) in feature space.
  • Figure 2: The comparison of generic model performance under different communications round between FedFRliu2022fedfr and our method. AdaFedFR achieves the outperforming performance in 2 communication rounds.