Artificial Intelligence Perspectives on Mexican Art: A Case Study

Abstract

In this article a set of images corresponding to paintings of eight painters considered an Artistic Heritage of Mexico was clustered to identify clusters of images with similar characteristics between themselves. The images were acquired from a public source available on the Internet, a Pre-processing phase was applied in order to standardize the images in size and number of pixels, an extraction phase of features was applied for each image using Principal Components Analysis (PCA) and Histograms of Oriented Gradients (HOG), a segmentation phase of the features that were derived in the extraction phase was applied using the K-Means technique and the quality of the clusters that were obtained was evaluated using the Silhouette measure. As a result, seven clusters were attained with interesting characteristics: two of the most renowned Mexican painters worldwide whose artistic work is known for using a rich variety of shapes and colors (Diego Rivera and Frida Kahlo) clearly predominated in two clusters; an artist who is recognized for capturing Mexican landscapes in his paintings (José María Velasco) predominated in another cluster; in other three clusters a mixture of various Mexican artists predominated and in the last cluster Diego Rivera clearly predominated. According to the results, it seems that the paintings of Diego Rivera stand out due to a greater number of shapes used compared to the rest of the paintings analyzed. This article is a sample of the potential of Artificial Intelligence applied to Mexican art (and to art in general).