aytacpacal / xaida4detection

XAIDA4Detection

Open source code for the detection and characterization of spatio-temporal extreme events

Description

The XAIDA4Detection toolbox consists of a full pipeline for the detection and characterization of extreme events using ML and automatic image processing tools. Its purpose is to provide an ML-based generic and flexible pipeline to detect and characterize extreme events based on spatio-temporal Earth and climate observational data. The pipeline consists of three different stages for both the detection and characterization of extreme events:

1. Data loading and pre-processing
2. ML architecture selection and training
3. Evaluation and visualization of results

Usage

# 1) Create an empty pip environment
python3 -m venv ./xaida_env 


# 2) Activate environment
source ./xaida_env/bin/activate


# 3) Install dependencies
pip install -r requirements_xaida.txt install libs


# 4) Run main.py of XAIDA4Detection using a config file. Some examples:

# DroughtED database and K-Nearest Neighbors (KNN) model (from PyOD) 
python main.py --config=/configs/config_DroughtED_PYOD.yaml

# DroughtED database and LSTM-based architecture (user-defined) 
python main.py --config=/configs/config_DroughtED_LSTM.yaml

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

Please make sure to update tests as appropriate.

Citation

If you use this code for your research, please cite XAIDA4Detection: A Toolbox for the Detection and Characterization of Spatio-Temporal Extreme Events:

Cortés-Andrés, J., Gonzalez-Calabuig, M., Zhang, M., Williams, T., Fernández-Torres, M.-Á., Pellicer-Valero, O. J., and Camps-Valls, G.: XAIDA4Detection: A Toolbox for the Detection and Characterization of Spatio-Temporal Extreme Events, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-4816, https://doi.org/10.5194/egusphere-egu23-4816, 2023.

Acknowledgement

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement 101003469.

License

MIT