Natural Stone Image Classification Using Online Databases and Convolutional Neural Networks

ABSTRACT: In this paper, we discuss the relevance of two distinct types of online natural stone databases (technical-institutional repositories and logistics-commercial e-platforms) for extracting (training and testing) input images and creating an automatic visual inspection system for natural stone classification. Using convolutional neural networks (CNNs) and images from a selected online data repository, a Deep Learning (DL) system was developed to estimate the class of the natural stone in a given image. The DL models were developed through transfer learning from existing image classifiers, as pre-trained classifiers were retrained on our dataset. Our best model achieved an Accuracy of 70.3% and an F-score of 0.67 for 70 classes.