top of page
AI Mineral Tech
AUTOMATIC ANALYSIS OF
ROCK THIN SECTIONS
Set of mineral labels

Image characteristics
For each rock type (diorites, gabbros, granites, granodiorites and syenites) 80 images were considered, 40 in PPL and 40 in XPL.
A total of 400 images with dimensions of 960×600 px in PNG (RGB) format for the training and validation set, plus a set of 20 images with the same characteristics for the evaluation.
The rocks are fresh or slightly altered and the thin sections have an average thickness of 30 μm.


Mineral categories
Eight minerals were labeled: potassium feldspar (Afs), amphibole (Amp), biotite (Bt), muscovite (Ms), olivine (Ol), pyroxene (Px), plagioclase (Pl), and quartz (Qz).
The abbreviations of Whitney & Evans (2010) were used.
A total of 62,178 labels were obtained
bottom of page