2021 年 20 巻 2 号 p. 95-104
Drawing is an important skill, but it is difficult for people who are weak in visual or spatial perception. In this study, we aim at utilizing computer support to assist people’s visual perception during their drawing practice. To this end, we propose to use AI technology based on CNN (convolutional neural network) model to recognize what people draw, and then output CG (computer graphics) according the recognition result. We apply this proposed solution to support the practice in creating design patterns like the tessellation. Our work includes four contributions in-below. (1) recognition of the user’s original drawing based on our trained AI model; (2) generation of new visual design by incorporating the result of recognition into the pattern called tessellation; (3) a digital interface to interactively provide the drawing recognition and tessellation generation; (4) our work is the first trial to realize the innovation tetra proposed by Dr. H. Shiizuka. A cross validation shows the theoretically accuracy of the train AI model. The result of a questionnaire-based evaluation shows the practical usability of the support system, and also advise us future improvement to reach various user requirements.
