---
license: cc-by-nc-4.0
task_categories:
- image-classification
- other
tags:
- climate
- geospatial
- remote-sensing
- earth-science
- satellite-imagery
- sar
- time-series
- disaster-response
- earthquake-detection
pretty_name: QuakeSet
size_categories:
- 1K<n<10K
---
# QuakeSet

QuakeSet is a dataset to analyze different attributes of earthquakes. It contains bi-temporal time series of images and ground truth annotations for magnitudes, hypocenters, and affected areas.

- **PrePrint:** https://arxiv.org/abs/2403.18116
- **Published Verions:** https://ojs.iscram.org/index.php/Proceedings/article/view/10

You can find an official implementation on [TorchGeo](https://torchgeo.readthedocs.io/en/stable/api/datasets.html#non-geospatial-datasets).

## Dataset Details

### Dataset Description

The images are taken from the Sentinel-1 mission using the Interferometric Wide swath mode.

The International Seismological Centre provides information about earthquakes.

- **License:** cc-by-nc-4.0

## Dataset Structure

The dataset is divided into three folds with equal distribution of magnitudes and balanced in positive and negative examples.

Each sample contains:
- an image
- x,y coordinates
- epsg of the coordinates
- affected label
- magnitude

*NOTE*: the main file is *earthquakes.h5*, while *p_earthquakes.h5* contains the private set for the [SMAC Challenge](https://www.codabench.org/competitions/2222/). 
This set contains only placeholder ground truth. 

## Citation

**BibTeX:**
```
@article{Rege_Cambrin_2024,
   title={QuakeSet: A Dataset and Low-Resource Models to Monitor Earthquakes through Sentinel-1},
   ISSN={2411-3387},
   url={http://dx.doi.org/10.59297/n89yc374},
   DOI={10.59297/n89yc374},
   journal={Proceedings of the International ISCRAM Conference},
   publisher={Information Systems for Crisis Response and Management},
   author={Rege Cambrin, Daniele and Garza, Paolo},
   year={2024},
   month=may
}
```