Deep-Learning-Based Automatic Mineral Grain Segmentation and Recognition

A multitude of applications in engineering, ore processing, mineral exploration, and environmental science require grain recognition the counting minerals. Typically, this task is performed manually with drawback monopolizing both time resources. Moreover, it requires highly trained personnel a wealth knowledge equipment, such as scanning electron microscopes optical microscopes. Advances machine learning deep make possible to envision automation many complex tasks various fields at an accuracy equal human performance, thereby, avoiding placing resources into tedious repetitive tasks, improving efficiency, lowering costs. Here, we develop deep-learning algorithms automate minerals directly from grains captured Building upon our previous work applying state-of-the-art technology, modify superpixel segmentation method prepare data for algorithms. We compare two residual network architectures (ResNet 1 ResNet 2) classification identification processes. achieve validation 90.5% using 2 architecture 47 layers. Our approach produces effective application while also achieving better rate than reported thus far literature process other well-known, deep-learning-based models, including AlexNet, GoogleNet, LeNet.