Glacier Mapping with Deep Learning

Authors:
Konstantin A. Maslov,
Thomas Schellenberger,
Claudio Persello

A repository for the teaching materials on deep learning for glacier mapping from the Machine Learning for Glaciology workshop, Finse Alpine Research Centre, Norway. This part of the workshop covers glacier mapping with the use of pixel-based methods such as random forests and deep learning models for semantic image segmentation.

Materials

• Jupyter notebook on data preparation with CREODIAS and Python geospatial libraries
• Jupyter notebook on random forest for glacier mapping
• Jupyter notebook on deep learning for glacier mapping
• Introduction to machine learning slides
• Random forest and deep learning for glacier maping slides

Alternatively, you can access the materials on Google Drive, it includes large files (e.g. pretrained ResUNet) which we had to omit because of the GitHub limitations. The dataset can be downloaded here.

We will update the materials during the workshop. If you notice any inaccuracies, mistakes or errors, feel free to submit a pull request.