"Location-aware Adaptive Normalization: A Deep Learning Approach for Wildfire Danger Forecasting" by Mohamad Hakam Shams Eddin, Ribana Roscher and Juergen Gall. Published in IEEE Transactions on Geoscience and Remote Sensing

For conda, you can install dependencies using yml file:

  conda env create -f environment.yml

or using requirements.txt:

  conda create --name LOAN --file requirements.txt

For pip:

  pip install -r requirements.txt

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

The dataloader for FireCube dataset:

  FireCube_dataloader.py

For training:

  train.py

For testing:

  test.py

To train on FireCube dataset, You can download the training/testing samples from https://zenodo.org/record/6528394 (~250GB).

Compress the zip file of the datasets.tar.gz and copy the file mean_std_train.json into the directory datasets/datasets_grl/npy/spatiotemporal

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

Pretrained models can be downloaded from pretrained_models

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

@ARTICLE{LOAN,
  author={Shams Eddin, Mohamad Hakam and Roscher, Ribana and Gall, Juergen},
  journal={IEEE Transactions on Geoscience and Remote Sensing}, 
  title={Location-Aware Adaptive Normalization: A Deep Learning Approach for Wildfire Danger Forecasting}, 
  year={2023},
  volume={61},
  number={},
  pages={1-18},
  doi={10.1109/TGRS.2023.3285401}}

  
@article{LOAN,
  title={Location-aware Adaptive Denormalization: A Deep Learning Approach For Wildfire Danger Forecasting},
  author={Mohamad Hakam Shams Eddin and Ribana Roscher and Juergen Gall},
  journal={ArXiv},
  year={2022},
  volume={abs/2212.08208}}

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

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