Multi-Modal Foundation · isp-uv-es
ML4Floods
Optical satellite image segmentation model for global flood extent mapping from Sentinel-2 and Landsat data.
What ML4Floods is
ML4Floods is an image segmentation framework developed by isp-uv-es and researchers Enrique Portalés-Julià, Gonzalo Mateo-García, Cormac Purcell, and Luis Gómez-Chova. It provides automated flood delineation from optical satellite imagery to support earth science research, climate analysis, disaster response, and humanitarian operations.
Mapping inundation rapidly and accurately is crucial during flood emergencies. ML4Floods addresses this need by providing cloud-aware segmentation models trained to extract flood extents across diverse global environments from satellite passes.
How it works
The system is built on multioutput Unet neural network architectures. Pre-trained model variants are configured for different sensor inputs and spectral combinations, including a full-band configuration, an RGB and near-infrared (RGBNIR) configuration, and a cross-sensor configuration designed to transfer between Sentinel-2 and Landsat-8 using visible and shortwave infrared bands.
The models process optical satellite scenes to perform multioutput, cloud-aware segmentation, differentiating surface water and flood extents even when clouds or atmospheric interference are present.
Key facts
- Architecture
- Unet multioutput
- Supported sensors
- Sentinel-2 and Landsat (optical)
- Primary task
- Image segmentation (flood extent mapping)
- Release year
- 2023
What you can use it for
- Near-real-time flood extent mapping
- Segmenting water coverage in recently acquired Sentinel-2 or Landsat optical passes to evaluate inundation boundaries during disaster events.
- Cross-sensor satellite analysis
- Applying models across Sentinel-2 and Landsat-8 data using aligned visible and shortwave infrared band combinations.
- Limited-band optical segmentation
- Extracting flood boundaries from platforms providing only red, green, blue, and near-infrared spectral channels.
Run it here
We list this model but cannot run it yet: no library and no architectures — nothing can load it.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Provides multioutput Unet variants adapted to different spectral band combinations, including full spectral inputs and standard RGBNIR bands.
- Includes cross-sensor model configurations capable of processing both Sentinel-2 and Landsat scenes.
- Features cloud-aware segmentation designed to handle atmospheric challenges in optical remote sensing.
Watch out for
- Operates exclusively on optical satellite imagery and cannot map floods beneath opaque cloud cover without sensor alternatives.
- Model variants require specific spectral band selections (such as RGBNIR or shortwave infrared bands) matching the respective pre-trained configuration.
Inputs and outputs
You provide
Image
PNG, JPEG or TIFF. Sent to the model as-is.
You get back
The returned image.
Labels with confidence scores, highest first.
Licence
Check the licence before commercial use · cc-by-nc-4.0
The cc-by-nc-4.0 licence carries terms worth reading before you depend on it.
Questions
Can I use ML4Floods commercially?
Check the licence before commercial use. The cc-by-nc-4.0 licence carries terms worth reading before you depend on it.
Which satellite imagery sources are supported?
The models are designed to segment optical imagery collected by Sentinel-2 and Landsat satellite missions.
What model variants are available?
The available pre-trained models comprise an all-band multioutput Unet, an RGBNIR multioutput Unet, and a Sentinel-2 to Landsat-8 model based on visible and shortwave infrared bands.
How does the model handle clouds?
The system uses cloud-aware segmentation to account for cloud presence when identifying flood extents in optical scenes.
How to cite it
E. Portalés-Julià, G. Mateo-García, C. Purcell, and L. Gómez-Chova. Global flood extent segmentation in optical satellite images. Scientific Reports 13, 20316 (2023). DOI: 10.1038/s41598-023-47595-7.
Written from ML4Floods's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.