A Sentinel-2 benchmark dataset for deep-learning active-fire segmentation across 25 California wildfires, July 2020-August 2026
Abstract
This article describes an open image dataset for developing and evaluating active-fire segmentation methods in satellite imagery. The dataset contains 2,148 image-mask pairs from 25 California wildfires, with acquisitions spanning July 2020 to August 2026. Each image is a 512x512-pixel, three-channel composite derived from Sentinel-2 Level-2A bands B12, B11 and B8A at 20 m spatial sampling. A fixed linear rendering is applied throughout the dataset. Corresponding masks distinguish background, SWIR-rule active fire and invalid observations. The masks were generated from shortwave-infrared brightness and near-infrared contrast, followed by constrained neighborhood growth. The release includes chip-level metadata and an incident-disjoint partition containing 18 training, three validation and four test fires. Among the image pairs, 841 contain active-fire labels; these labels occupy 0.0766% of all grid cells. A mask-blind analyst review covers 233 test chips and provides a separate assessment of the rule-generated labels at chip and connected-component levels. Reference training and evaluation code accompanies the data, including a ResNet-34 U-Net implementation with validation-based checkpoint and threshold selection. The archived images, masks, metadata and review annotations support research on rare-class segmentation, learning from algorithmic labels and transfer across fire incidents. The versioned dataset is deposited on Zenodo, with preparation and reuse software maintained in a public GitHub repository.
Versions
Related
- Evaluating Cross-Country Generalization of Deep Learning-Based Flood Mapping Using Sentinel-1 SAR and Sentinel-2 Imagery
- Multi-temporal Canopy Height Modeling using Deep Learning and Sentinel-2: The Influence of Training Data and Input Feature Representations
- Estimating aman and aus rice area in Bangladesh using sentinel-1 imagery and machine learning algorithms
- Sentinel-1 SAR time series for forest phenology: multimodal deep learning with meteorological fusion across deciduous broadleaf and evergreen needleleaf forests
- Forecasting the moisture content of live forest fuels with a transformer machine learning model and Sentinel-2 satellite data
BibTeX
@misc{mitra2026sentinel2,
title = {{A Sentinel-2 benchmark dataset for deep-learning active-fire segmentation across 25 California wildfires, July 2020-August 2026}},
author = {Shreyan Mitra and Mohammadreza Narimani and Parastoo Farajpoor},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2609.16199},
eprint = {2609.16199},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.16199},
}
RIS
TY - UNPB
TI - A Sentinel-2 benchmark dataset for deep-learning active-fire segmentation across 25 California wildfires, July 2020-August 2026
AU - Shreyan Mitra
AU - Mohammadreza Narimani
AU - Parastoo Farajpoor
PY - 2026
DA - 2026/10/24
PB - arXiv
DO - 10.48550/arxiv.2609.16199
UR - https://arxiv.org/abs/2609.16199
AB - This article describes an open image dataset for developing and evaluating active-fire segmentation methods in satellite imagery. The dataset contains 2,148 image-mask pairs from 25 California wildfires, with acquisitions spanning July 2020 to August 2026. Each image is a 512x512-pixel, three-channel composite derived from Sentinel-2 Level-2A bands B12, B11 and B8A at 20 m spatial sampling. A fixed linear rendering is applied throughout the dataset. Corresponding masks distinguish background, SWIR-rule active fire and invalid observations. The masks were generated from shortwave-infrared brightness and near-infrared contrast, followed by constrained neighborhood growth. The release includes chip-level metadata and an incident-disjoint partition containing 18 training, three validation and four test fires. Among the image pairs, 841 contain active-fire labels; these labels occupy 0.0766% of all grid cells. A mask-blind analyst review covers 233 test chips and provides a separate assessment of the rule-generated labels at chip and connected-component levels. Reference training and evaluation code accompanies the data, including a ResNet-34 U-Net implementation with validation-based checkpoint and threshold selection. The archived images, masks, metadata and review annotations support research on rare-class segmentation, learning from algorithmic labels and transfer across fire incidents. The versioned dataset is deposited on Zenodo, with preparation and reuse software maintained in a public GitHub repository.
ER -