This dataset is related to the manuscript “Context-Aware Deep Learning Ensembles for Beach-Surface and Grain-Size Classification in Isbjørnhamna, Svalbard”.
It contains 15 maps of beach-surface and 15 maps of grain-size classification at 1 m resolution for the Isbjørnhamna beach. The maps are based on an existing set of digital elevation models and orthomosaics at 0.05 m resolution created using DJI Phantom 4 Pro surveys and Structure-from-Motion photogrammetry over 2018-2024 (https://doi.org/10.5281/zenodo.19327624), grain size data collected in the field in 2022-2024 (https://doi.org/10.5281/zenodo.21128398) and a deep-learning classification model.
The beach-surface classes include:
0. Snow
1. Dry sediment
2. Wet sediment
3. Water
4. Bedrock
5. Boulder
6. Driftwood
7. Algae
8. Ice
9. Anthropogenic
In the grain-size set, classes 1 (Dry sediment) and 2 (Wet sediment) are re-classified to:
11. Fine
12. Medium
13. Coarse
The other classes are the same as in the beach-surface classification.