Generating a pixel-wise annotated training dataset to train ML algorithms for mineral identification in Mineral thin sections

Published: 3 July 2020| Version 1 | DOI: 10.17632/cp897djgn9.1
Contributor:
Jiaxin Yu

Description

The dataset contains two folders. 1. In the "virtual_microscopy" folder, you can find a short screen recording showing how PetroScan works. PetroScan is the built-in toolbox of Virtual Petrography (ViP) platform used for displaying digital thin section images under different modes. The video records the extinction behavior and the change of inference color of quartz and mica in a mylonite sample under cross-polarized when moving the angle selection wheel. 2. "Superpixel_evaluation_example" contains boundary maps and raw images of a subarea of Bentheimer Sandstone. The superpixel segmentation results using different superpixel algorithms can be found in the subfolder "qualitative_evaluation".

Files

Institutions

Rheinisch Westfalische Technische Hochschule Aachen

Categories

Mineral

License