Geometric Knowledge-Assisted Federated Dual Knowledge Distillation Approach Towards Remote Sensing Satellite Imagery

TL;DR

This paper introduces a novel federated learning framework that leverages geometric knowledge and dual knowledge distillation to improve remote sensing satellite imagery analysis, effectively handling data heterogeneity across satellites.

Contribution

The proposed GK-FedDKD framework innovatively combines geometric knowledge with dual knowledge distillation to enhance federated learning for satellite imagery analysis.

Findings

Outperforms state-of-the-art methods by 68.89% on EuroSAT dataset.

Uses geometric knowledge for embedding augmentation to improve model training.

Demonstrates robustness across multiple remote sensing datasets.

Abstract

Federated learning (FL) has recently become a promising solution for analyzing remote sensing satellite imagery (RSSI). However, the large scale and inherent data heterogeneity of images collected from multiple satellites, where the local data distribution of each satellite differs from the global one, present significant challenges to effective model training. To address this issue, we propose a Geometric Knowledge-Guided Federated Dual Knowledge Distillation (GK-FedDKD) framework for RSSI analysis. In our approach, each local client first distills a teacher encoder (TE) from multiple student encoders (SEs) trained with unlabeled augmented data. The TE is then connected with a shared classifier to form a teacher network (TN) that supervises the training of a new student network (SN). The intermediate representations of the TN are used to compute local covariance matrices, which are aggregated at the server to generate global geometric knowledge (GGK). This GGK is subsequently employed for local embedding augmentation to further guide SN training. We also design a novel loss function and a multi-prototype generation pipeline to stabilize the training process. Evaluation over multiple datasets showcases that the proposed GK-FedDKD approach is superior to the considered state-of-the-art baselines, e.g., the proposed approach with the Swin-T backbone surpasses previous SOTA approaches by an average 68.89% on the EuroSAT dataset.