應用類神經網路於電腦輔助流行眼鏡設計之研究

Application of Neural Network to Computer Aided Fashion Glasses Design

鄭智文(Chih-Wen Cheng)

指導教授 ： 林家旭

台南應用科技大學/生活科技學院/生活應用科學研究所/碩士(2015年)

摘要

近年來，產品的生命週期明顯較以往縮短，再加上電腦輔助設計軟體的技術進步，因此，產品的構想階段導入自動化設計概念，遂成為設計的重要議題，藉此獲得大量、快速的設計構想，來激發設計師的創造力。
研究目的即為建立一套應用類神經網路學習功能之眼鏡造型設計的輔助平台。
研究方法首先選取12支流行眼鏡款為樣本，以20位眼鏡專家進行流行太陽眼鏡設計屬性與意象認知評估。進而將進行眼鏡意象認知與設計元素之關係評量，作為訓練類神經網路流行眼鏡設計屬性與意象語意的設計規則，並利用倒傳遞類神經網路演算結果建立電腦輔助應用程式。
研究結果利用語意分析法建立的流行眼鏡意象設計模式，歸納出「高雅的」、「前衛的」、「成熟的」與「熱情的」四個影響人們對於眼鏡意象的心理重要因素，可輔助產品設計師找出滿足消費者產品意象之最佳造型要素組合，以提昇整體產品造型意象之效用；以3D模型作為調查，應用語意分析法歸納流行眼鏡意象認知與設計元素之間的關係，期望能掌握眼鏡元素與意象營造的設計規則，作為類神經網路學習之依據樣本，以建立流行眼鏡設計的人工智慧；類神經網路與電腦輔助軟體相結合，進行類神經網路分類學習，意即擁有類似人類所具備的生物神經網路功能，可利用其對流行眼鏡意象認知的評選能力，來建立電腦的類神經網路評選機制與標準。
本研究提供設計者能快速精確地掌握消費者心理層面上特定的意象需求，並且轉換為產品的設計意象與造型要素，並運用此眼鏡造型設計的輔助軟體，未來在眼鏡造型設計的過程中，創造出更多、更有創意的眼鏡造型概念。

關鍵字

眼鏡造型、流行意象、類神經網路

並列摘要

In recent years, the product life cycle is shorter than in the past. With the progress of the computer-aided design technology, in the process of create product implement the concepts of automated design, has become an important issue. Furthermore, the creativity of the designers will be stimulated a large number of rapid design concepts. The research objective is to apply the learning function of neural network to establish an auxiliary software of Fashion Glasses Design.
The methods are to select twelve popular sunglasses in the beginning as the samples, which evaluated the popular glasses design attributes and the imagery of cognitive by the twenty glasses experts. Then, find the relationship between the amount of glasses assessment cognitive image and design elements as training the neural network design rules. Use the calculation results of the back propagation neural network to establish computer-aided application popular eyewear designs.
The results are using semantic analysis method to establish the glasses image design patterns. The mode concludes are "Elegant", "Avant-garde", "Mature" and "Passionate" four factor which affect people for mental imagery important factor ,which can assist product designers to find the best combination which meet the consumer the imagery of cognitive that enhance overall's effectiveness. Use the 3D model as surveys, then application the semantic analysis induction the relationship between popular imagery cognition and glasses design elements. Hope can master the glasses design elements and imagery cognition to create design rules, as a sample of neural network based learning, then create artificial intelligence. Neural networks and computer-aided software combined to conduct neural network classify, which means it have human-similar biological neural networks function, the ability to use their vote to the popular perception of the image of glasses, to create a computer class neural network with the standard selection mechanism.
This study provides the designer can quickly and accurately grasp the specific needs of the consumer psychological imagery and converted into product design image and form factor. In the future, using the glasses design auxiliary application software can help that create more and more creative concepts shape glasses at the glasses shape design process.

並列關鍵字

glasses style, fashion image, Neural Networks

參考文獻

外國期刊:

Chan, C. S., (1992). Exploring individual style in design. Environment and Planning B: Planning and Design, 19, 503-523.

Crilly, N., Moultrie, J., and Clarkson, P. J., (2004). Seeing things: consumer response to the visual domain in product design. Design Studies, 20, 1-31.

Frank Rosenblatt., (1957). The perceptron: A Probabilistic model for Visual Perception, Process of the International Congress of Psychology, North Holland, 15, 290-297.

Hopfield, J. J., (1982). Neural networks and physical systems with emergent collective computational abilities. Proc. Natl. Acad. Sci USA, 79, 2554-2558.