Automatic Classification of Plutonic Rocks with Deep Learning

  • Germán Harvey Salinas
  • Elías Lael Vázquez Espinosa
Palabras clave: Plutonic Rocks, Computer Vision, Machine Learning, Neuronal Networks, Mobile Application, Android


Igneous rocks form from cooling and solidication of molten rock, either from magma within the Earth’s crust (plutonic rocks), or from lava extruded on to the Earth’s surface in the atmosphere or underwater (volcanic rocks). The classification of igneous rocks can be done using data from X-ray uorescence (XRF), neutron activation analysis (INAA and RNAA), inductively coupled plasma emission spectrometry (ICP), atomic absorption spectrophotometry (AAS), and mass spectrometry. However, these approaches tend to be expensive and dicult to understand and apply. In this study, several models for the classication of plutonic rocks were created with a convolutional neural network developed with TensorFlow. Specifically, several combinations of gabbro, granite, and granodiorite image samples were used in the experiments, each class with 97 images. The best result was obtained with images of granite and granodiorite rocks. The evaluation of this model was satisfactory, with an accuracy value of 73.03% and average precision, recall, and F1 score values of 86 %. An Android mobile application uses this classification model to return the class of a rock image. This mobile application oers the exibility
to carry out rock classications in the eld in real time.

Cómo citar
Salinas, G. H., & Vázquez Espinosa, E. L. (2020). Automatic Classification of Plutonic Rocks with Deep Learning. anuario2020, 1(1), 170-176. Recuperado a partir de