In recent decades, wildfires, as widespread and extremely destructive natural disasters, have caused tremendous property losses and fatalities, as well as extensive damage to forest ecosystems. Many fire risk assessment projects have been proposed to prevent wildfires, but GIS-based methods frequently suffer from inflexible data features and limited generalizability. Inspired by the abundance of publicly available remote sensing projects and the burgeoning development of deep learning in computer vision, our research focuses on assessing fire risk using remote sensing imagery.
In this work, we propose a novel remote sensing dataset, FireRisk, consisting of 7 fire risk classes with a total of 91,872 labelled images for fire risk assessment. This remote sensing dataset is labelled with the fire risk classes supplied by the Wildfire Hazard Potential (WHP) raster dataset, and remote sensing images are collected using the National Agriculture Imagery Program (NAIP), a high-resolution remote sensing imagery program.
This figure shows sample images of all 7 labels in our FireRisk dataset. The images measure 270 × 270 pixels, with a total of 91,872 image.
Dataset | Model | pre-trained | Accuracy | F1-score | Precision | Recall |
---|---|---|---|---|---|---|
FireRisk | ResNet-50 | ImageNet1k | 63.20 | 52.56 | 52.75 | 53.41 |
FireRisk | ViT-B/16 | ImageNet1k | 63.31 | 52.18 | 53.91 | 51.15 |
FireRisk | DINO | ImageNet1k | 63.36 | 52.60 | 54.95 | 51.27 |
FireRisk | DINO | UnlabelledNAIP | 63.44 | 52.37 | 53.79 | 51.75 |
FireRisk | MAE | ImageNet1k | 65.29 | 55.49 | 56.42 | 55.36 |
FireRisk | MAE | UnlabelledNAIP | 63.54 | 52.04 | 54.09 | 51.78 |
50% FireRisk | ResNet-50 | ImageNet1k | 62.09 | 50.27 | 51.07 | 50.41 |
50% FireRisk | ViT-B/16 | ImageNet1k | 62.22 | 50.07 | 52.20 | 50.15 |
50% FireRisk | DINO | ImageNet1k | 61.75 | 51.21 | 51.35 | 51.63 |
50% FireRisk | DINO | UnlabelledNAIP | 62.49 | 51.35 | 52.08 | 51.48 |
50% FireRisk | MAE | ImageNet1k | 63.70 | 50.23 | 52.85 | 51.94 |
50% FireRisk | MAE | UnlabelledNAIP | 62.68 | 52.05 | 52.63 | 51.59 |
20% FireRisk | ResNet-50 | ImageNet1k | 61.37 | 49.53 | 50.28 | 50.12 |
20% FireRisk | ViT-B/16 | ImageNet1k | 61.43 | 48.80 | 50.89 | 48.53 |
20% FireRisk | DINO | ImageNet1k | 60.95 | 50.72 | 50.99 | 51.28 |
20% FireRisk | DINO | UnlabelledNAIP | 61.96 | 50.83 | 53.03 | 50.62 |
20% FireRisk | MAE | ImageNet1k | 62.51 | 51.13 | 52.46 | 50.87 |
20% FireRisk | MAE | UnlabelledNAIP | 61.80 | 50.07 | 51.69 | 49.11 |
From the table we can draw the following conclusions:
The maximum accuracy for the supervised benchmarks can reach 63.31%, while for the self-supervised benchmarks, the MAE pre-trained on ImageNet1k can achieve the optimal accuracy of all models at 65.29%. And the checkpoint of the optimal model is MAE pre-trained on ImageNet1k
Our self-supervised learning benchmarks outperform supervised learning on FireRisk, although their improvement on less training data is limited.
Our new pre-trained latent representations have a considerable increase in DINO, which can reach 63.44% compared to 63.36% for the DINO pre-trained on ImageNet.
FireRisk: A Remote Sensing Dataset for Fire Risk Assessment with Benchmarks Using Supervised and Self-supervised Learning
Image naming in our FireRisk: $(pointid)_(grid_code)_(x_coord)_(y_coord).png$
Name | Data Type | Meaning |
---|---|---|
FID | integer | ID of the data point in the file |
pointid | integer | unique ID of the data point in the WHP dataset |
grid_code | integer(from 1 to 7) | code for fire risk level |
class_desc | string(seven values) | description of the fire risk level, corresponding to the grid_code, which are 1:Very Low, 2:Low, 3:Moderate, 4:High, 5:Very High, 6:Non-burable and 7:water |
x_coord | number | longitude coordinates of the grid centroid |
y_coord | number | latitude coordinates of the grid centroid |
DINO | MAE | |
---|---|---|
pre-trained checkpoint | download | download |
If you have used our FireRisk dataset, please cite the following papers: https://arxiv.org/abs/2303.07035
@misc{shen2023firerisk,
title={FireRisk: A Remote Sensing Dataset for Fire Risk Assessment with Benchmarks Using Supervised and Self-supervised Learning},
author={Shuchang Shen and Sachith Seneviratne and Xinye Wanyan and Michael Kirley},
year={2023},
eprint={2303.07035},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
This research was undertaken using the LIEF HPC-GPGPU Facility hosted at the University of Melbourne. This Facility was established with the assistance of LIEF Grant LE170100200.