Self-Adaptive Computational Aesthetic Evaluation of Chinese Ink Paintings Based on Deep Learning

Computational aesthetic evaluation of artworks has become an active research direction in recent years. However, current works mainly focus on oil paintings and photographs, there have been few attempts quantitative Chinese ink paintings. painting uses blended with water a variation brushwork to depict picture, which differs significantly from photographs visual features, semantic principles. Aiming at this problem, we propose framework self-adaptive computational based deep learning technique. Firstly, build standard dataset for images. Secondly, according principles paintings, design multi-view parallel neural network by taking global images local patches as multi-column inputs extract features. Finally, model subject query mechanism. Experimental results show that, compared the basic VGG16 architecture, our that contains six paralleled convolution groups higher performance, features outperform traditional hand-crafted proposed can predict human decision highly significant Pearson correlation 0.823, mean squared error 0.161. Moreover, interference experiments is sensitive factors including composition, color, texture. Our work not only offers deeply-learned-based reference but also reveals relationship between perceptions deeply-learned extracted