Design of Japanese Kimono (Yukata) using an Interactive Genetic Algorithm

Transcription

1 THE SCIENCE AND ENGINEERING REVIEW OF DOSHISHA UNIVERSITY, VOL. 50, NO. April 2009 Design of Japanese Kimono (Yukata) using an Interactive Genetic Algorithm Maiko SUGAHARA * Mitsunori MIKI ** and Tomoyuki HIROYASU *** (Received January 20, 2009) In recent years, the design of yukata changed from the fixed traditional designs to various designs. People are interested in the design of yukata. It is useful to design a yukata suitable for each preference. But, in many cases, people have ambiguous image for their favorite yukata, it is difficult to make their favorite design. We propose a yukata design system using an Interactive Genetic Algorithm (). The proposed system is for designing a yukata to suit user s taste. From the assessment experiment of the system, it was found that the proposed system proved to be effective in the designing of a yukata. In addition, we proposed additional functions that allow obi (sash) color mutation partially in search for the solution. And the experimental results showed the effectiveness of the additional functions. Key words Optimization, Interactive Genetic Algorithm, Yukata design system, Color combination., 2) * Graduate Student, Department of Knowledge Engineering and Computer Sciences, Doshisha University, Kyoto Telephone: , Fax: , msugahara@mikilab.doshisha.ac.jp ** Department of Knowledge Engineering and Computer Sciences, Doshisha University, Kyoto Telephone: , Fax: , mmiki@mail.doshisha.ac.jp *** Faculty of Life and Medical Sciences,Doshisha University, Kyoto Telephone: , Fax: , tomo@is.doshisha.ac.jp 6

2 7 Fig. Interactive Genetic Algorithm 3) Alternative solutions 3-D CG 4), 5) 6) User Evaluate Display System GA Fig.. system RGB HSB 2. HSB 8) HSB (Genetic Algorithm:GA) 7)

3 HSB Fig. 2 HSB Fig Fig. 2 number H S B H S B H S B Yukata fabric Obi Select favorite design Yukata fabric Obi Pattern gene Yukata fabric number Pattern number (0:plane,:stripe) (0-23) Individual Pattern Userinterface for selection of first Individu- Fig. 4. als Fig. 2. Chromosome. 2 Fig Fig. 3 Fig. Continue seach button End seach button Generation count Start Evaluation tool * Button * Slider bar Generation of first individual Display Evaluation Human Operation Fig. 5. Example of display. Terminal Criterion Yes End No Selection Crossover Mutation 5 Fig. 3. Flow chart of yukata design system. 8

4 n n n=2 2 2 BLX-α 9) BLX-α 2 α 0 Fig. 6. A Fig. 6.. B 2. 2 BLX-α d Example of a crossover for kimono fabric's Hue. ParentA 4.2 ParentA d Offspring Fig. 7 Fig. 7 B ParentB d Range of generating offsprings ParentB Offspring Fig. 6. Example of crossover. N V N V Fig. 7. Example of final design. Fig. 8 Fig. 9

5 20 8 5% 9% 86% Clear image Can somewhat imagine and express in words Cannot express in words but can somewhat imagine Can't imagine at all What kind of yukata do you imagine wearing (imagine a girl wearing) to a fireworks show? How much can you imagine it? Fig. 8. Result of questionnaire item. Fig. 9 2 Fig. 9 Fig. 0 Change the color of obi 5% 5% 67% 23% Yes Somewhat designed it Can t really say Couldn t really design it No Were you able to design a yukata that fitted your design concept with this system? Change the obi color of only individuals that were stochastically selected. Fig. 9. Result of questionnaire item 2. Fig. 0. User interface after action of the button

6 2 0% Yes % 36% Somewhat designed it Can t really say Couldn t really design it No Were you able to design a yukata that fitted the concept by using the obi color mutation Fig.. Result of questionnaire item % 5% 9% 59% 8% Improved system Preferred improved system Can't really say Preferred basic system Basic system Which of the two systems (the basic and the improved) did you find easier to design a yukata that fitted the concept better with? Fig. 2. Result of questionnaire item Fig. 3 Fig Fig. Fig. 2 3 Fig. 4 Fig. 2 Fig

7 22 Subject Before After Final design Subject 2 Subject 3 Example of yukata by using improved sys- Fig. 3. tem. Elite individual Mutation individual ) T. Sano, H. Ukida, H. Yamamoto. Adaptive texture alignment for japanese kimono design. Proceedings of the IEEE Instrumentation and Measurement Technology Conference, Vol.2,pp. pp , ) T. Sano, H. Nagahata, H. Yamamoto. Design support system for japanese kimono. IECON 98 Proceedings of the 24th Annual Conference of the IEEE Industrial Electronics Society, Vol.,pp , ),, , , pp. 4) K. Aoki and H Takagi. 3-D CG Lighting with an Interactive GA. st Int l Conf. on Conventional and Knowlidge-based Intelligent Electronic Systems, pp , ) H.-S. Kim and S.-B. Application of interactive genetic algorithm to fashion design. Engineering Applications of Artificial Intelligence 3(6), Vol., pp , ),,,,,.., Vol. 0, No. 2, pp. 3 22,

8 23 7) D.E.Goldberg. Genetic Algorithms in Search Optimization and Machine Learnig. Addison- Wesley, ),.., ) L.J Eshleman and J.D Schaffer. Real-Coded Genetic Algorithms and Interval-Schemata. Foundations of Genetic Algorithms, Vol. 2, pp , 993. Fig. 7. Example of final design Change the color of obi Select favorite design Fig. 4. Userinterface for selection of first Individuals Change the obi color of only individuals that were stochastically selected. Fig. 0. User interface after action of the button Continue seach button End seach button Generation count Subject Before After Final design Evaluation tool * Button * Slider bar Subject 2 Fig. 5. Example of display Subject 3 d ParentA d Example of a crossover for kimono fabric's Hue. ParentA Offspring B ParentB Offspring d Range of generating offsprings ParentB Fig. 6. Example of crossover Elite individual Mutation individual Fig. 3. Example of yukata by using improved system 23