The purpose of this paper is to propose a new methodology that approaches the problem of art analysis by combining AI, especially the recent technology of GAN (Generative adversarial network), and psychological experiments. Furthermore, using the methodology, we tackled the task of clarifying the differences and similarities in impressions of Eastern/Western figurative/abstract art and the position of media art produced by one of the authors, Naoko Tosa. Several image sets have been prepared; a flower photo set (A), an art image set of Impressionism paintings (B1), an art image set of Western abstract paintings (B2), an art image set of Chinese representational paintings (B3), and an art image set of abstract images created by Naoko Tosa, one of the authors (B4). The transformation from A to each of B was carried out using GAN and thus four image sets, for which their original artist and art genres were anonymized, were obtained. Using the obtained four image sets, a psychological experiment was carried out by asking subjects consisting of 23 students to fill in questionnaires. By analyzing the obtained questionnaires, we have found the following. Abstract paintings and representational paintings are judged to be different. Representational paintings in the West and East were judged to be similar. Abstract images by Naoko were judged as similar to Western abstract images. These results show that AI could be used as an analysis tool to investigate differences among artworks and art genres without a bias caused by the prejudice most subjects have toward specific artists, artworks, or art genres.
View full abstract