Focal_TSMP

This is the code to reproduce the results presented in the paper:

"Focal-TSMP: Deep learning for vegetation health prediction and agricultural drought assessment from a regional climate simulation" by Mohamad Hakam Shams Eddin, and Juergen Gall

Computer Vision Group, Institue of Computer Science, University of Bonn

Project Page | Paper

Model architecture

Setup

For conda, you can install dependencies using yml file:

  conda env create -f Focal_TSMP.yml

Code

The code has been tested under Pytorch 1.12.1 and Python 3.10.6 on Ubuntu 20.04.5 LTS with NVIDIA RTX A6000 GPUs and NVIDIA GeForce RTX 3090 GPU.

The configuration file:

  config.py

The dataloader for TSMP dataset:

  TerrSysMP_NET_dataset.py

For training:

  train.py

For testing:

  test.py

Dataset

To train on TSMP dataset:

You have to download the preprocessed remote sensing data into the directoy data from https://zenodo.org/doi/10.5281/zenodo.10008814 (~6GB).
Download the TSMP simulation (~900GB) from Juelich Research Centre into the directory data/TerrSysMP. You can use the script download_data_Juelich.py
Generate the dataset TerrSysMP_NET using the script generate_TerrSysMP_NET.py.

The file (generate_TerrSysMP_NET.py) is hard coded for this specific dataset and not ment to be a generic script for preprocessing.

To train on another dataset, you need to create a new dataloader class like TerrSysMP_NET_dataset.py

Checkpoints

Pretrained models can be downloaded from https://zenodo.org/doi/10.5281/zenodo.10015048

Citation

If you find our work useful in your research, please cite:


@Article{egusphere-2023-2422,
AUTHOR = {Shams Eddin, M. H. and Gall, J.},
TITLE = {Focal-TSMP: Deep learning for vegetation health prediction and agricultural drought assessment from a regional climate simulation},
JOURNAL = {EGUsphere},
VOLUME = {2023},
YEAR = {2023},
PAGES = {1--50},
URL = {https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2422/},
DOI = {10.5194/egusphere-2023-2422}
}

Acknowledgments

This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) within the Collaborative Research Centre SFB 1502/1–2022 - DETECT - D05.

License

The code is released under MIT License. See the LICENSE file for details.

tfgbestneal/Focal_TSMP