This collection contains deep learning models (.h5) for sematic segmentation of explosive volcanic plumes. 3 trained models are available, using 3 different model architectures/backbones: efficientnetb0.h5: FPN architecture with the efficientnetb0 backbone inceptionv3.h5: UNet architecture with the inceptionv3 backbone densenet121.h5: UNet architecture with the densenet121 backbone Also include is a python file (utils.py) containing a few useful functions for working with these models. their use is explained in the docstrings within the file.