Skip to content

Latest commit

 

History

History
35 lines (20 loc) · 1.3 KB

README.md

File metadata and controls

35 lines (20 loc) · 1.3 KB

mesogeos Track B: Final Burned Area Prediction

Dataset

You have two options, you can either download the dataset or generate it yourself using the dataset_extraction.py script.

Downloading the data

TODO: Add instructions to download the dataset.

Generating the data

With the mesogeos datacube downloaded (see mesogeos/README.md), you can generate the dataset using the dataset_extraction.py script.

python dataset_extraction.py --ds_path <path_to_datacube> --gdf_path <path_to_shapefile> --save_dir <save_directory> --patch_half <half_of_patch_size_around_ignition> --days_before <days_before_ignition> --days_after <days_after_ignition>

Reproducing the experiments

Set up a conda environment with the environment.yml file. Disclaimer: this environment file is not minimal, it contains all the packages used during the development of the project. Also, it is not guaranteed to work on all platforms. It has been tested on Ubuntu 20.04 with an NVIDIA RTX 3080 GPU.

conda env create -f environment.yml

Set up wandb and fill appropriate values in the run script or disable wandb from src/main.py.

Run the script.

bash scripts/run_experiment.sh