https://github.com/isaaccorley/earthquake-monitoring-by-ranking

[ECML-PKDD 2024 MACLEAN] Estimating Earthquake Magnitude in Sentinel-1 Imagery via Ranking
https://github.com/isaaccorley/earthquake-monitoring-by-ranking

metric-learning sentinel-1 synthetic-aperture-radar

Last synced: about 1 month ago
JSON representation

[ECML-PKDD 2024 MACLEAN] Estimating Earthquake Magnitude in Sentinel-1 Imagery via Ranking

• Host: GitHub
• URL: https://github.com/isaaccorley/earthquake-monitoring-by-ranking
• Owner: isaaccorley
• License: mit
• Created: 2024-05-04T02:33:53.000Z (8 months ago)
• Default Branch: main
• Last Pushed: 2024-07-30T18:35:00.000Z (5 months ago)
• Last Synced: 2024-11-26T23:36:21.418Z (about 1 month ago)
• Topics: metric-learning, sentinel-1, synthetic-aperture-radar
• Language: Python
• Homepage:
• Size: 4.01 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        


Estimating Earthquake Magnitude in Sentinel-1 Imagery via Ranking


[**Daniele Rege Cambrin**](https://darthreca.github.io/)1 · [**Isaac Corley**](https://isaacc.dev/)2 · [**Paolo Garza**](https://dbdmg.polito.it/dbdmg_web/people/paolo-garza/)1 · [**Peyman Najafirad (Paul Rad)**](https://scholar.google.com/citations?user=uoCn8c8AAAAJ&hl=en)2

1Politecnico di Torino, Italy    2University of Texas at San Antonio, USA

**[ECML-PKDD 2024 MACLEAN Workshop](https://sites.google.com/view/maclean24/home)**







**In this paper, we propose to pose the estimation of earthquake magnitudes as a metric-learning problem, training models to not only estimate earthquake magnitude from Sentinel-1 satellite imagery but to additionally rank pairwise samples**. Our experiments show at max a 30%+ improvement in MAE over prior regression-only based methods, particularly transformer-based architectures.



    


    Figure 1. QuakeSet samples varied by earthquake magnitude. Each sample contains a pair of pre and post earthquake event Sentinel-1 (SAR) imagery containing 2 bands (VV & VH). The samples in this figure are plotted as false color images (VV, VH, VV/VH) along with their magnitudes.



### Install Dependencies

```bash

pip install -r requirements.txt

```

### Dataset

We utilize the [TorchGeo](https://github.com/microsoft/torchgeo) implementation of the [QuakeSet dataset](https://arxiv.org/abs/2403.18116) in our experiments. It can be downloaded automatically by running the following:

```python

from torchgeo.datasets import QuakeSet

root = "path/to/download/folder"

dataset = QuakeSet(root, download=True)

```

### Training

Make sure to edit the config files in the `configs/` folder to specify the timm backbone architecture and logger backend you desire to use. Then simply run

```bash

python train.py

```

### Evaluation

Given the path to the trained checkpoint run the following:

```bash

python test.py --checkpoint path/to/checkpoint.ckpt --device 0

```

## Citation

If you find this project useful, please consider citing:

```bibtex

@article{cambrin2024estimating,

  title={Estimating Earthquake Magnitude in Sentinel-1 Imagery via Ranking},

  author={Cambrin, Daniele Rege and Corley, Isaac and Garza, Paolo and Najafirad, Peyman},

  journal={arXiv preprint arXiv:2407.18128},

  year={2024}

}

```