Polygon files for the annotations in FLOGA v2. The annotations remain the same as in v1, but new imagery has been included adopting the re-processed Sentinel-2 products. In some cases, images from different dates have been used compared to v1 due to more accurate cloud masking, thus rejecting cloudy images that were previously deemed cloud-free and vice versa.
We have also added candidate Sentinel-1 imagery in case the user is interested to produce a dataset with a SAR modality. We opted to include 2 pre- and 2 post-fire Sentinel-1 products, useful for coherence map production. FLOGA is currently being extended with such data aiming to tackle cloud/smoke coverage and potentially real-time wildfire progression monitoring.
Each polygon file contains the following fields:
ID: unique identifier for each wildfire eventYear: The year of the eventStart date: The ignition date of the wildfireEnd date: The date the wildfire endedS2_s: The pre-fire Sentinel-2 image usedS2_e: The post-fire Sentinel-2 image usedMOD_s: The pre-fire MODIS image usedMOD_e: The post-fire MODIS image usedS1_s1: The first pre-fire Sentinel-1 imageS1_s2: The second pre-fire Sentinel-1 imageS1_e1: The first post-fire Sentinel-1 imageS1_e2: The second post-fire Sentinel-1 image
If you would like to use our work, please cite our paper:
@ARTICLE{10479972,
author={Sdraka, Maria and Dimakos, Alkinoos and Malounis, Alexandros and Ntasiou, Zisoula and Karantzalos, Konstantinos and Michail, Dimitrios and Papoutsis, Ioannis},
journal={IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing},
title={FLOGA: A Machine-Learning-Ready Dataset, a Benchmark, and a Novel Deep Learning Model for Burnt Area Mapping With Sentinel-2},
year={2024},
volume={17},
number={},
pages={7801-7824},
doi={10.1109/JSTARS.2024.3381737}
}