"Human Supervision and Control in Engineering and Music"

Sung-Bae Cho 

Inspiration-based Media Retrieval with Interactive Genetic Algorithm

Abstract

Evolutionary computation has shown a great potential to work out several real-world problems in the point of optimisation, but it is still quite far from realizing a system of matching the human performance, especially in creative applications. To overcome this shortcoming, we present a promising technique called interactive genetic algorithm (IGA), which performs optimisation with human evaluation and the user can obtain what he has in mind through repeated interaction with. To show the usefulness of the IGA to develop effective inspiration-based systems, we have applied it to the problems of fashion design and emotion-based image retrieval.
 

Interactive Genetic Algorithm

First proposed by John Holland in 1975, genetic algorithm (GA) as one of the computational implementations is an attractive class of computational models that mimic natural evolution to solve problems in a wide variety of domains. However, most of the conventional applications of GA lack of the capability to utilize human intuition and emotion appropriately in creative applications such as architecture, art, music, and design. There is no clear measure to give the evaluation of fitness other than the one in the human mind.

Interactive GA (IGA) is a technique that performs optimization with the human evaluation. A human can obtain what he has in mind through repeated interaction with the method, when the fitness function cannot be explicitly defined. This allows us to develop effective human-oriented evolutionary systems, since this obtains from human the fitness value for the problem at hand, and produces better designs or images for the next generation.
 

Application to Fashion Design 

Fig. 1 shows the fashion design aid system developed based on the IGA. There is a database of partial design elements, which are stored as 3D models. The system selects the models of each part and combines them into a number of individual designs. The population is displayed on screen and user gives fitness values to each design. Then, the system reproduces the population proportional to the fitness value of each design, and applies crossover and mutation to make the next generation. The results are displayed again in the screen with 3D graphics. Iteration of these processes can produce the population of higher fitness value, namely better designs.

Fig.1: Fashion design system using IGA
 

Application to Image Retrieval

The system is constructed as shown in Fig. 2. In the preprocessing step, at first, wavelet transform is performed for every image in the database and stored are the overall average color and the indices and signs of the m magnitude wavelet coefficients in a search table. The system displays twelve images, obtains the fitness values of the images from human, and selects candidates based on the fitness. Genetic operation, vertical or horizontal crossover, is applied to the selected candidates. To find the next twelve images, the stored image information is evaluated by each criterion. Twelve images of the higher magnitude value are provided as a result of the search.

Fig. 2: Image retrieval system using IGA
 

Concluding Remarks

We have presented an approach that implements inspiration-based media manipulation systems with human preference and emotion using interactive genetic algorithm. Several experiments show that our approach allows to design and search digital media not only explicitly expressed image, but also abstract images such as “cheerful impression,” “gloomy impression,” and so on. It is expected that the same approach can be applied to many problems in music retrieval and manipulation based on intuition and inspiration.
 

References

Cho, S.-B., Lee, J.-Y. (2001). A human-oriented image retrieval system using interactive genetic algorithm, IEEE Trans. Systems, Man and Cybernetics-Part A. (in press)

Kim, H.-S., Cho, S.-B. (2000). Application of interactive genetic algorithm to fashion design, Engineering Applications of Artificial Intelligence, (13) 635-644.

Takagi, H. (1998). Interactive evolutionary computation: Cooperation of computational intelligence and human KANSEI, Proc. of Int. Conf. on Soft Computing, 41-50.