COLOR COMBINATION ADAPTATION TO HUE SPREAD INDEX USING AN AUTOENCODER

This study introduces a method for adapting color combinations using an Autoencoder (AE) model. In design, color combination adaptations often follow established schemes of color harmony, such as monochromatic, analogous, complementary, and triadic schemes, all of which are predominantly influenced by hue. To quantify the spread of hue on the hue wheel, this study proposes the Hue Spread Index (HSI), a numerical metric that represents the different color schemes. The training dataset includes color combinations and their corresponding HSI values, which are integrated through the AE model to establish a reciprocal relationship between them. After training the AE model, a series of color generation experiments were conducted. The results reveal that adjusting the HSI value can affect the spread of hue on the hue wheel: increasing the HSI value leads to a wider spread of colors, while decreasing the HSI value results in a narrower spread. These findings demonstrate that the AE model developed in this study successfully establishes a correspondence between color schemes and HSI values. It provides a practical and innovative approach to meeting designers' needs for automated and flexible color adaptation, advancing the application of machine learning in color design.