Mineral texture classification using deep convolutional neural networks: an application to zircons from porphyry copper deposits

The texture and morphology of igneous zircon indicates magmatic conditions during zircon crystallisation and can be used to constrain provenance. Zircons from porphyry copper deposits are typically prismatic, euhedral and strongly oscillatory zoned which may differentiate them from zircons associated with unmineralised igneous systems. Here, cathodoluminesence images of zircons from the Quellaveco porphyry copper district, Southern Peru, were collected to compare zircon textures between the unmineralised Yarabamba Batholith and the Quellaveco porphyry copper deposit. Quellaveco porphyry zircons are prismatic, euhedral and strongly oscillatory zoned, whereas the batholith zircons contain more variable morphologies and zoning patterns. We adopt a deep convolutional neural networks (CNNs) approach to demonstrate that a machine can classify porphyry zircons with a high success rate. We trial several existing CNN architectures to classify zircon images: LeNet-5, AlexNet and VGG, including a transfer learning approach where we used the weights of a VGG model pre-trained on the ImageNet dataset. The VGG model with transfer learning is the most effective approach, with accuracy and ROC-AUC scores of 0.86 and 0.93, indicating that a Quellaveco porphyry zircon CL image can be ranked higher than a batholith zircon with 93% probability. Visualising model layer outputs demonstrates that the CNN models can recognise crystal edges, zoning and mineral inclusions. We trial implementing trained CNN models as unsupervised feature extractors, which can empirically quantify crystal textures and morphology. Therefore, deep learning provides a powerful tool for the extraction of petrographic information from minerals which can be applied to constrain provenance in detrital studies.