Quantifying shape descriptors for aesthetic concepts

Abstract

Usually, consumers refer to products using adjectives or personalities commonly associated to humans. For example, by saying “this is a cute mug” or “this is an aggressive car”. This way of referring to products leads to what has been called the “product personality”. As described in [Govers 2004], the “product personality” can be defined by a set of characteristics that people use to describe and discriminate products from others. The product personality can be based on the shape of the product, but also its colour, texture, smell and even the sound can be relevant to define the product personality [Janlert 1997]. Examples of 23 product personalities can be found in [Ortiz 2011]. It has been demonstrated that the product personality perceived by consumers have direct relation with some visual product aesthetic features like simplicity, harmony, balance, unity, dynamics, novelty and timeliness/fashion [Brunel 2007], all features well understood by industrial designers as well as artists. In this sense, usually the effective achievement of a determined personality on a product shape depends mainly on the designer personal experience and skills. This work is oriented on developing a systematic and quantitative understanding about how a specific concept (like personality) can be imprinted in a product by modifying its appearance. Because consumers usually prefer products which personality is similar to their own [Govers 2005], the deep understanding on how a specific personality can be introduced in the appearance of a product under development, can be important knowledge for successfully introduce new products in the market. Our ongoing research seeks to build such knowledge and to make available to designers and engineers, a flexible and customizable new tool that allows the exploration of a range of formal solutions, by applying geometric modifiers which imprint concepts intuitively and smoothly (like personality) to products under shape development. This article presents the preliminary results of a long term research that aims to develop such new tool. While in [Prieto, 2013] the authors describe the problem in general terms and started analyzing the correlation among shapes and perception, this paper focus on the identification of features and measurement methods to quantitatively assess product personality. At this stage, a first approach of shape descriptor allowing to properly categorizing the object shapes according to one pair of opposite concepts (Gentle-Aggressive) is presented. A survey has been conducted to obtain statistical measurements of the perception degree of the interviewees with respect to the concept pair Gentle-Aggressive, for each image subject to this study. Then, the descriptor and the measurement (statistical mean or median) for each object silhouette are utilized in a supervised learning process to obtain a model of the concept. The obtained model corresponds to an artificial neural network, which allows to predict the perception degree of the Gentle-Aggressive concept from the descriptor of a new silhouette. In the following, a summary of advances on supporting CAD tools on early stages of the design process and geometric shape descriptors is presented. Then the first preliminary shape descriptor is described and validated.