Quantitative evaluation of sensitivity in confidential car exterior design

Abstract

In recent years, the manufacturing industry has seen a shift in competition from performance, which can easily be evaluated numerically, to design which much more challenging to express numerically. The rise of companies that focus on design, such as Apple, Samsung and IKEA, is remarkable. However, design presents two challenges for the manufacturing industry. First, the sensory aspect of design is challenging to evaluate quantitatively, and unified evaluation indicators are not yet defined. Second, confidentiality of product design. In many cases, the design is kept in confidence within the companies, so it is often hesitated to investigate large customers. The above two problems increase the influence of the evaluator's experience and cause a situation that it is challenging to create a design desired by the customer. Therefore, the present study aims to enable inexpensive quantitative evaluation of automobile exterior design while maintaining confidentiality. We propose a technique that uses a convolutional neural network to link features extracted from accumulated design images to the sensitivity extracted from the customer's voice. This is then used to quantitatively evaluate an input image.