Prototype for automated damage detection in Sentinel 1 imaging of iceselves using unsupervised learning.
The prototype makes use of a VAE architecture, and employs a bespoke input genertation framework to create cutouts for network training from larger imaging tiles.
Notebooks have been used for intial prototyping. These have been converted to scripts for a first prouction prototype.