Awesome-Remote-Sensing-Foundation-Models

Awesome Remote Sensing Foundation Models

The Awesome Remote Sensing Foundation Models project is a comprehensive, ever-evolving repository that consolidates an array of academic papers, datasets, benchmarks, and pre-trained model weights specifically designed for Remote Sensing Foundation Models (RSFMs). This project acts as a valuable resource for researchers, developers, and enthusiasts involved in remote sensing, providing them with the tools and information they need to advance their work.

Latest Updates

The project is actively maintained, with the latest update noted on October 30, 2024. Recent updates include new survey papers and advancements in models like Change-Agent, PANGAEA, TEOChat, SAR-JEPA, and PIS.

Table of Contents

The project is organized into several sections:

• Models

  • Remote Sensing Vision Foundation Models
  • Remote Sensing Vision-Language Foundation Models
  • Remote Sensing Generative Foundation Models
  • Remote Sensing Vision-Location Foundation Models
  • Remote Sensing Vision-Audio Foundation Models
  • Remote Sensing Task-specific Foundation Models
  • Remote Sensing Agents

• Datasets & Benchmarks

  • Benchmarks for RSFMs
  • (Large-scale) Pre-training Datasets

• Others

  • Relevant Projects
  • Survey Papers

Remote Sensing Vision Foundation Models

This section highlights vision foundation models in remote sensing, detailing various projects with their abbreviations, titles, publication details, linked papers, and, when available, access to their code and pre-trained weights. For instance:

• GeoKR: Geographical Knowledge-Driven Representation Learning for Remote Sensing Images, published in TGRS 2021. Interested users can find the paper here and access the code here.

• GASSL: Geography-Aware Self-Supervised Learning, showcased in ICCV 2021. The model explores self-supervised learning tailored to geographical awareness, with details and code here.

• SeCo: Seasonal Contrast leverages unsupervised pre-training with uncurated remote sensing data, as seen in ICCV 2021. Users can review the research here and utilize the code here.

Each model listing provides insights into the innovative techniques being employed to advance remote sensing technologies. These models reflect various research themes from geographical knowledge incorporation to self-supervised learning methodologies, emphasizing the diversity and breadth of research in the RSFM field.

Conclusion

The Awesome Remote Sensing Foundation Models project is a central hub that connects the remote sensing community with critical developments and resources. By presenting this collection, the project supports the ongoing evolution of remote sensing technologies, offering stakeholders the means to harness, explore, and innovate within the space.