AI-BASED TEXTURE DESIGN SYSTEM BASED ON AGE THINKING MODEL

抄録

In this study, we propose a system that automatically generates various texture images for the purpose of design concepts for a large number of users and customize production for individual users using the AGE Thinking Model and Deep Convolutional GAN. The AGE Thinking Model is used to determine the development guidelines for the system. Among the three types of thinking in design, AI is in charge of generation and analysis, whereas humans are in charge of evaluation. The proposed system has two types of evaluation systems: one for a large number of users' sensitivities, which screens training images using a sensitivity model constructed from texture image features and sensory evaluation values, and one for individual users' sensitivities that screens training images by having each user evaluate the training images interactively. The results of applying the system to a wood-grain texture design showed that the system produced images that improved the sensory evaluation values of “high-grade,” which had little variation among users, and “preference,” which had a large variation.