Published May 29, 2026
| Version v1
Software
Open
Trained models for mineral phase segmentation in SEM images of clastic rocks - supplementary material (Deliverable 2.1)
Authors/Creators
-
1.
Fraunhofer Research Institution for Energy Infrastructures and Geothermal Systems
-
2.
Geological Survey of Denmark and Greenland
Description
SEM-classifier: Pre-trained models for Mineral Segmentation in SEM images of clastic rocks.
Provided are pre-trained machine- and deep-learning models for autmated segmentation of
mineral phases in Scanning Electron Microscope (SEM) images of clastic sandstone. Training was
performed on a dataset provided by GEUS.
The models classify pixels in paired sets of BSE (Backscattered Electron) and CL (Cathodoluminescence)
images into four phases / classes: Quartz, Overgrowth, Porespace, Other mineral.
See README.md for more information.