The OPARA service was recently upgraded to a new technical platform. You are visiting the outdated OPARA website. Please use https://opara.zih.tu-dresden.de/ for new data submissions. Previously stored data will be migrated in near future and then the old version of OPARA will finally be shut down. Existing DOIs for data publications remain valid.
Manually delineated glacier calving fronts of 23 Greenland and 2 Antarctic outlet glaciers from 2013 to 2021 and source code for automated extraction by deep learning
![No Thumbnail [100%x80]](data:image/svg+xml;base64,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)
Abstract
Supplementary material to the manuscript: Loebel, E., Scheinert, M., Horwath, M., Heidler, K., Christmann, J., Phan, L., Humbert, A., Zhu, X. (2022): Extracting glacier calving fronts by deep learning: the benefit of multi-spectral, topographic and textural input features, IEEE Transactions on Geoscience and Remote Sensing.
The dataset is composed of three parts. Firstly, this dataset provides 728 manually delineated glacier calving front positions of 23 Greenland and 2 Antarctic outlet glaciers from 2013 to 2021. For that, we provide ocean masks and frontal positions as Polygon or LineString Shapefiles respectively. For manual delineation, we used optical Landsat-8 imagery. Secondly, we provide 995 input raster subsets (17 layers each) with their corresponding, manually delineated, segmentation masks. Input raster subsets have dimensions of 512 pixels by 512 pixels and are available in both png and georeferenced tif format. These machine learning ready raster subsets are optimized for training and validating artificial neural networks. Thirdly, we lay out our source code we use for neural network training and accuracy assessment. The code uses python3 and shell scripts.