Skip to content

The 1st VHR SAR-Optical benckmark dataset for detecting earthquake damaged buildings.

License

Notifications You must be signed in to change notification settings

ya0-sun/PostEQ-SARopt-BuildingDamage

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

20 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

QQB: QuickQuakeBuildings

This repo contains dataset, code, and explanation of the paper: QuickQuakeBuildings: Post-earthquake SAR-Optical Dataset for Quick Damaged-building Detection by Yao Sun, Yi Wang, and Michael Eineder.

👉 Check out the LinkedIn Article for the background and motivation.

Introduction

Quick and automated earthquake-damaged building detection from post-event satellite imagery is crucial, yet it is challenging due to the scarcity of training data required to develop robust algorithms. In this work, we provide the first dataset dedicated to detecting earthquake-damaged buildings from post-event very high resolution (VHR) Synthetic Aperture Radar (SAR) and optical imagery.

Utilizing open satellite imagery and annotations acquired after the 2023 Turkey-Syria earthquakes, we deliver a dataset of coregistered building footprints and satellite image patches of both SAR and optical data, encompassing more than four thousand buildings. The task of damaged building detection is formulated as a binary image classification problem, that can also be treated as an anomaly detection problem due to extreme class imbalance. We provide baseline methods and results to serve as references for comparison.

Datasource

Satellite imagery

  • SAR

    The SAR image was obtained from Capella Space Synthetic Aperture Radar Open Dataset.

    The used image is of type Geocoded Terrain Corrected (GEO), and downloaded from here.

  • Optical

    The optical image was obtained from Maxar Analysis-Ready Data (ARD) under Maxar’s open data program. The CATALOG ID and QUAD KEY of the used imagery are listed below.

    CATALOG ID QUAD KEY
    1040010082698700 031133012123
    031133012132
    031133012301
    031133012310

Building footprints and labels of destroyed buildings

Post-event building footprints and labels of destroyed buildings were obtained from OpenStreetMap and Humanitarian OpenStreetMap Team:

Usage

Downaload the dataset from here, and then extract the dataset to ./data. The dataset should be organized as follows:

data
├── damaged
│   ├── OSMID_SAR.mat
│   ├── OSMID_SARftp.mat
│   ├── OSMID_opt.mat
│   ├── OSMID_optftp.mat
└── |── ...
│   intact
│   ├── ...
└── fold-1.txt
└── fold-2.txt
└── fold-3.txt
└── fold-4.txt
└── fold-5.txt

We use ImageNet weights for optical images and footprints, and SAR-HUB weights for SAR imagery, which were pretrained on TerraSAR-X data. You can download the weights here.

Depending on the training mode, run:

# sar + opt
python main.py \
--root ./data \
--val_split fold-1.txt \
--mode all \ # choose from [all, sar, opt]
--checkpoints checkpoints_all/rn18_pretrain_fold1 \
--sar_pretrain ./weights/ResNet18_TSX.pth \
--opt_pretrain imagenet \
# sar
python main.py \
--root ./data \
--val_split fold-1.txt \
--mode sar \ # choose from [all, sar, opt]
--checkpoints checkpoints_all/rn18_pretrain_fold1 \
--sar_pretrain ./weights/ResNet18_TSX.pth \
# opt
python main.py \
--root ./data \
--val_split fold-1.txt \
--mode opt \ # choose from [all, sar, opt]
--checkpoints checkpoints_all/rn18_pretrain_fold1 \
--opt_pretrain imagenet \

Results

Benchmark results on the dataset.

Examples of prediction results using SAR, optical, and both data, from their corresponding best model.

Acknowledgment

We would like to thank Capella Space and Maxar Technologies for providing satellite imagery under CC BY 4.0, and OpenStreetMap and Humanitarian OpenStreetMap Team for providing building footprints and annotations of destroyed buildings under ODbL License.

Citation

If you find the repo useful, please consider cite the following paper:

@article{sun2023qqb,
  author={Sun, Yao and Wang, Yi and Eineder, Michael},
  journal={IEEE Geoscience and Remote Sensing Letters}, 
  title={QuickQuakeBuildings: Post-Earthquake SAR-Optical Dataset for Quick Damaged-Building Detection}, 
  year={2024},
  volume={21},
  pages={1-5},
  doi={10.1109/LGRS.2024.3406966}}

About

The 1st VHR SAR-Optical benckmark dataset for detecting earthquake damaged buildings.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages