# ml4floods pre-trained models

This repository contains the trained models of the publication:

> E. Portalés-Julià, G. Mateo-García, C. Purcell, and L. Gómez-Chova [Global flood extent segmentation in optical satellite images](https://www.nature.com/articles/s41598-023-47595-7). _Scientific Reports 13, 20316_ (2023). DOI: 10.1038/s41598-023-47595-7.

We include the trained models:
* Unet multioutput - `models/WF2_unetv2_all`
* Unet multioutput S2-to-L8 - `models/WF2_unetv2_bgriswirs`
* Unet multioutput RGBNIR - `models/WF2_unetv2_rgbi`


![metrics_ml4floods](metrics_ml4floods.png)

In order to run any of these models in Landsat or Sentinel-2 scene see the tutorial [*Inference with clouds aware floods segmentation model*](https://spaceml-org.github.io/ml4floods/content/ml4ops/HOWTO_Run_Inference_multioutput_binary.html) in the ml4floods docs.

---
license: cc-by-nc-4.0
---