2023 Volume 22 Issue 1 Pages 1-10
When creating a new product design, a designer chooses design solutions from concepts that combine different impression words. However, there could be more than one optimal design solution. Therefore, a reverse inference system that returns all optimal design solutions is necessary. The purpose of this study is twofold: First, using the eyeglass design as a case study, we clarify the capability to search for diverse design solutions using deep generative models, a solution method for reverse inference. Second, we compare differences in the explored design solutions between two deep generative models, CVAE and JMVAE. The two models generated diverse eyeglass frame designs and yielded results having different characteristics in the search range.