Distinguishing between Abstract Art by Artists vs. Children and Animals

Abstract

expressionism is a school of art characterized by nonrepresentational paintings where color, composition, and brush strokes are used to express emotion. These works are often misunderstood by the public who see them as requiring no skill and as images that even a child could have created. However, a recent series of studies has shown that ordinary adults untrained in art or art history, as well as young children, can differentiate paintings by abstract expressionists and superficially similar works by preschool children and even animals (monkeys, apes, elephants). Adults perform this distinction with an accuracy rate of ∼64%, significantly higher than chance. Here we ask whether machine perception can do as well. Using the same paintings, we show that in ∼68% of the cases the computer algorithm can discriminate between abstract paintings and the work of children and animals. We also applied a method that computes the correlation between the degree of artisticity deduced from human perception of the paintings and the visual content of the images, and we show significant correlation between perceived artisticity and visual content. The image content descriptor that was the strongest predictor of correct identification was the fractality of the painting. We also show that the computer algorithm predicts the perceived intentionality of the paintings by humans. These results confirm perceptible differences between works by abstract expressionists and superficially similar ones by the untrained and show that people see more than they think they see when looking at abstract expressionism.

Predicting and grouping digitized paintings by style using unsupervised feature learning

“…Shamir et al (53) have shown that computers are able to exceed in accuracy the detection of abstract art created by children versus actual artists. In this study, Shamir et al (53) used their specialized feature extraction algorithm (15,54), however, it may be possible to obtain similar results using the simple UFLK method demonstrated in our paper. Another similar use of the UFLK features could be to detect when an artist decides to change their painting style.…”

What S in a Name Experiment on the Aesthetic Judgments of Art Produced by Artificial Intelligence

“…Most of the researches carried out results using abstract paintings (Alvarez et al, 2015;Nissel et al, 2015;Shamir et al, 2016;Bubić et al, 2016;Mullennix & Robinet, 2018). Sometimes abstract paintings may be confusing to understand at first quick look, but if done correctly, they also have a profound appeal to them.…”

“…Therefore, differences could be identified not only by adults without any knowledge of art history but also by computers. One of the latest studies (Shamir et al, 2016), the researcher tried to find out if machine perception could distinguish differences between human, child, and animal abstract paintings. They show that the computer algorithm can discriminate between abstract paintings and children's and animal's work in 68 per cent of cases.…”

“…The museum in Oxford held its first solo exhibition of eight paintings by "Ai-Da", a humanoid robot with artificial intelligence (Stock, 2019). Nevertheless, one of the latest researches showed that creative machines could determine not only their creativity but also the creative results of other agents (Shamir et al, 2016).…”

A Deep Approach for Classifying Artistic Media from Artworks

“…The style matrix is used for image and style retrieval on image database. Shamir et al [17] presented a recognition scheme for abstract expressionism by evaluating the relation between the level of artistic and visual contents. Lecoutre et al [12] employed deep residual network for recognizing artistic style of artwork images in WikiArt database.…”