Summary of Bridging Data Islands: Geographic Heterogeneity-aware Federated Learning For Collaborative Remote Sensing Semantic Segmentation, by Jieyi Tan et al.
Bridging Data Islands: Geographic Heterogeneity-Aware Federated Learning for Collaborative Remote Sensing Semantic Segmentation
by Jieyi Tan, Yansheng Li, Sergey A. Bartalev, Shinkarenko Stanislav, Bo Dang, Yongjun Zhang, Liangqi Yuan, Wei Chen
First submitted to arxiv on: 14 Apr 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper tackles the challenge of remote sensing semantic segmentation in the face of geographic information security concerns. Federated learning, a privacy-preserving distributed collaborative learning technology, is proposed as a solution to leverage isolated remote sensing data across institutions. However, existing FL methods neglect the significant heterogeneity exhibited by remote sensing images from different institutions. The proposed Geographic heterogeneity-aware Federated learning (GeoFed) framework consists of three modules: Global Insight Enhancement (GIE), Essential Feature Mining (EFM), and Local-Global Balance (LoGo). GeoFed alleviates class distribution heterogeneity through a prior global class distribution vector, object appearance heterogeneity through essential features, and enables the model to possess both global generalization capability and local adaptation. Experimental results on three public datasets demonstrate that GeoFed consistently outperforms state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about finding a way to share remote sensing data across different places while keeping it private. Right now, this kind of data is often stored separately because of concerns about security and competition between institutions. The authors suggest using a special type of learning called federated learning, which allows people to work together on a project without sharing their individual data. However, the existing methods for doing this don’t take into account the fact that remote sensing images from different places look very different. To solve this problem, the authors propose a new framework that includes three main parts: making sure the class labels are consistent across all the data, finding the most important features in each image, and balancing the global and local aspects of the model. The results show that their approach is better than what’s currently available. |
Keywords
* Artificial intelligence * Federated learning * Generalization * Semantic segmentation
