SSL4Eco
🌍🌱

Official implementation for
SSL4Eco: A Global Seasonal Dataset for Geospatial Foundation Models in Ecology
CVPR EarthVision workshop 2025


If you ❤️ or simply use this project, don't forget to give the repository a ⭐, it means a lot to us !

@article{plekhanova2025ssl4eco,
  title={SSL4Eco: A Global Seasonal Dataset for Geospatial Foundation Models in Ecology},
  author={Plekhanova, Elena and Robert, Damien and Dollinger, Johannes and Arens, Emilia and Brun, Philipp and Wegner, Jan Dirk and Zimmermann, Niklaus},
  journal={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year={2025},
}

📌 Description

SSL4Eco is a Sentinel-2 dataset for pretraining geospatial foundation models. More specifically, this project proposes a recipe for building pretraining sets that capture the geographical and phenological diversity of ecosystems across the globe. We observe that this simple spatiotemporal sampling yields significant improvements on various downstream macroecological tasks.

📰 Updates

• 18.06.2025 💾 Released SSL4Eco and downstream task datasets as well as weights for our SeCo-Eco model
• 24.04.2025 🚧 Datasets, code, and weights will soon be publicly released !
• 11.06.2025 🌱 First code release

💻 Environment requirements

This project was tested with:

• Linux OS
• NVIDIA A100

The code may work in other environments but has not been thoroughly tested yet.

🏗 Installation

Simply run:

pip install -r requirements.txt

🔩 Project structure

└── ssl4eco
    ├── data_download             # For downloading SSL4Eco or downstream datasets 
    ├── docs                      # Project webpage
    ├── downstream_tasks          # For evaluating models on downstream tasks
    ├── index_files               # Metadata for SSL4Eco and our newly downstream datasets
    ├── pretraining               # For pretraining SeCo-Eco or MoCo-Eco on SSL4Eco
    ├── .gitignore                # List of files ignored by git
    ├── LICENSE                   # Project license
    ├── README.md                 # Readme
    └── requirements.txt          # Dependencies for pip install

🚀 Usage

Downloading datasets

See the data download section for further details on downloads.

Downloading pretrained weights

The weights for our SeCo-Eco model pretrained on the SSL4Eco dataset are available on huggingface 🤗.

Evaluation on downstream tasks

See the downstream tasks section for further details on evaluating foundation models on macroecological downstream tasks.

Pretraining

See the pretraining section for pretraining our SeCo-Eco or MoCo-Eco models on SSL4Eco.

Citing our work

If your work uses a part of the present code or ideas, please include the following citation:

@article{plekhanova2025ssl4eco,
  title={SSL4Eco: A Global Seasonal Dataset for Geospatial Foundation Models in Ecology},
  author={Plekhanova, Elena and Robert, Damien and Dollinger, Johannes and Arens, Emilia and Brun, Philipp and Wegner, Jan Dirk and Zimmermann, Niklaus},
  journal={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year={2025},
}

You can find our paper 📄 on arxiv.

Also, if you ❤️ or simply use this project, don't forget to give the repository a ⭐, it means a lot to us !