Prediction of Human Color Emotion on the Images Using Convolutional Neural Network

Abstract

In this study, we propose a method for predicting color emotion in real images using the mutual effect of color schemes. Features that take into account color interactions (synergy, contrast, area concentration, visual saliency, and perspective effects) are extracted from the image.　For example, synergy is the effect that the more warm colors contains in the image, the warmer it feels. On the other hand, the contrast is the effect that both the warm and cold colors contain in the image, the warmer it feels.　These features are then used to predict the color emotion of the color scheme by a convolutional neural network.　Then, we collect color emotion data for brightness and warmth through pairwise comparison experiments. We evaluate the effectiveness of the proposed method on the collected pairwise comparison data.