Application of Deep Learning in Petrographic Coal Images Segmentation

The study of the petrographic structure medium- and high-rank coals is important from both a cognitive utilitarian point view. constituents their individual characteristics features are responsible for properties coal way it behaves in various technological processes. This paper considers application convolutional neural networks images segmentation. U-Net-based model segmentation was proposed. network trained to segment inertinite, liptinite, vitrinite. segmentations prepared manually by domain expert were used as ground truth. results show that inertinite vitrinite can be successfully segmented with minimal difference liptinite turned out much more difficult segment. After usage transfer learning, moderate obtained. Nevertheless, image successful. good enough consider method supporting tool experts everyday work.