Abstract
This paper focuses on impression-based fabrication, a framework of personal fabrication that automatically generates 3D printing data of multiple designs suitable for user's desired impression, and develops a design system for picture frames as a case study. The framework consists of a design generation unit creating design candidates and an impression evaluation model assessing the candidates in terms of their appropriateness to the desired impression. In developing the system, an impression evaluation experiment is first conducted to clarify the relationship between various picture frames and their impression. By training neural networks using the experimental result, a set of impression evaluation models is implemented. The design generation unit is built based on the approach of a genetic algorithm in which the implemented model set is used as an evaluation function. An experiment evaluating the designs generated by the system demonstrates the validity of the framework in terms of the suitability and diversity of the designs.
