ibm-geospatial-saltmarsh-uk-rgbdem-extent-2m
This is a fine-tuned geospatial foundation model for detecting saltmarsh habitat in England using multi-spectral satellite imagery.
The model identifies the
extent of saltmarsh habitats from aerial imagery and was trained on labelled English
saltmarsh extents from the Environment Agency. The model was fine-tuned from
TerraMind, a general-purpose geospatial foundation model developed by
ESA and IBM.
Training Data
This model was trained using:
1. Saltmarsh Ground Truth Data
-
-
The model was trained on 256x256 labelled tiles, extracted at 10m resolution, where all saltmarsh zones (Pioneer, Spartina, Mid-Low, Upper, Reedbeds and Unclassified marsh) were merged to denote marsh extent.
-
The final training dataset size was train: 3539, val: 1156, test: 461 (total: 5156).
-
Saltmarshes in North West England were held out as the test set.
2. Aerial Imagery
-
-
The model was trained using 3 bands: RGB
3. Elevation
-
-
Elevation was provided at 2m resolution as an additional modality, as supported by Terramind.
How to Use
Example workflows and tutorials are available via the ML4EO workshop repository:
Acknowledgements
This work was supported by the Hartree National Centre for Digital Innovation, a collaboration between STFC and IBM.
Model Card Authors
Anne Jones, Helen Tamura-Wicks, Rosie Lickorish & Geoffrey Dawson.
IBM Public Repository Disclosure:
All content in this repository including code has been provided by IBM under the associated open source software license and IBM is under no obligation to provide enhancements, updates, or support. IBM developers produced this code as an open source project (not as an IBM product), and IBM makes no assertions as to the level of quality nor security, and will not be maintaining this code going forward.
Data Attribution
© Environment Agency copyright and/or database right 2023. Licensed under the Open Government Licence v3.0.
