Automatic classification of plutonic rocks with deep learning

Abstract

Igneous rocks form when molten magma is cooled and solidified, either within the Earth's crust (plutonic rocks), or from lava extruded onto the Earth's surface in the atmosphere or underwater (volcanic rocks). The classification of igneous rocks can be done using data from different instrumental techniques. However, these approaches tend to be expensive and time-consuming. In this research work, several models for the classification of granitoids, which are the most abundant plutonic rocks in the Earth's crust, were created with a convolutional neural network developed with TensorFlow. Specifically, several combinations of gabbro, diorite, tonalite, granodiorite, monzodiorite, and granite image samples were used in the experiments. The best result was obtained in the model that classifies images of gabbro, diorite, granodiorite, and granite with an accuracy value of 95%, an average precision value of 96%, an average recall value of 95%, and an average F1 score value of 95%.