Grant-in-Aid for Transformative Research Areas (A) Analysis and synthesis of deep SHITSUKAN information in the real world


D01-6 Quantitative Verification of Image Features Determining the Attractiveness of Painted Artwork Using Artificial Intelligence and Reverse Correlation Method


Tomoyuki Naito Osaka University

This project aims to verify the validity of the mental template hypothesis in judging the attractiveness of art by using the adversarial generative network (GAN) and the reverse correlation method, and to visualize the mental templates of attractive paintings as a high-quality image. In addition, by analyzing the feature vectors of the mental templates of attractive paintings, we quantitatively evaluate the image features that produce aesthetic sensation, and quantitatively evaluate the image features that produce universal pictorial attractiveness common to many people and the image features related to pictorial attractiveness with large individual differences.
Furthermore, by using the mental templates of attractive paintings obtained from 100 subjects, we will create new artworks that are highly attractive to individuals and groups using GAN, and verify whether these artworks can actually be highly evaluated in public art exhibitions.