michaeltrs / DeepSatData

Automatically create machine learning datasets from satellite images

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

DeepSatData: Building large scale datasets of satellite images for training machine learning models

plot DeepSatData is a toolkit for making datasets from satellite imagery suitable for training machine learning models. The process is split into two distinct parts:

  • identifying and downloading relevant Sentinel products for an area and time period of interest. Read more in download
  • processing downloaded products into datasets. Read more in dataset.

Further details on the methodology used can be found in our papers ["DeepSatData: Building large scale datasets of satellite images for training machine learning models"](arxiv url) and "Context-self contrastive pretraining for crop type semantic segmentation".

Dependencies

Install dependencies using pip

pip install -r requirements.txt

or creating a conda environment

conda create --name <env_name> --file requirements.txt

Citation

If you use DeepSatData in your research consider citing the following BibTeX entries:

@misc{tarasiou2021deepsatdata,
      title={DeepSatData: Building large scale datasets of satellite images for training machine learning models}, 
      author={Michail Tarasiou and Stefanos Zafeiriou},
      year={2021},
      eprint={2104.13824},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

@misc{tarasiou2021contextself,
      title={Context-self contrastive pretraining for crop type semantic segmentation}, 
      author={Michail Tarasiou and Riza Alp Guler and Stefanos Zafeiriou},
      year={2021},
      eprint={2104.04310},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

License

This project is under the CC-BY-NC 4.0 license. See LICENSE for details.

About

Automatically create machine learning datasets from satellite images

License:Other


Languages

Language:Python 64.1%Language:Jupyter Notebook 32.0%Language:Shell 3.9%