Development of Feature Extraction System for WADA Kunibo's works Focusing on Color

Abstract

This study proposes a system to extract the characteristics of the works of Kunibo Wada, an artist from Kagawa Prefecture, Japan, using image analysis technology. It is difficult for young designers to learn the style of Kunibo’s works of art simply by visiting art museums and viewing his works. Therefore, we have developed a system that helps young designers to understand the style of Kunibo’s works more easily than they have been able to do in the past. The system analyzes Kunibo’s style using artificial intelligence and helps young designers to understand his style by classifying works with similar styles. In particular, we developed an algorithm to extract features related to accent colors and "Kasumi" a technique for emphasizing objects. The results of evaluation experiments showed that both the precision rate and recall rate were sufficient for young designers to understand the style, which was the objective of the project. We expect that young designers will be inspired by the style of existing works of art to transmit culture from a new perspective, and to redesign and develop their works across fields.