Artificial intelligence technology is undergoing rapid development, and its function in art and design has evolved from a basic auxiliary tool to a core driving force for innovation. This study focuses on three key aspects in art and design: style transfer, composition optimization, and color matching, and thus constructs a hybrid optimization framework. This system integrates various technologies such as generative adversarial networks, reinforcement learning, and evolutionary algorithms, and incorporates user preferences and artistic aesthetic principles through the establishment of multidimensional evaluation criteria, thereby achieving automated iteration and continuous improvement of design schemes. The experimental results show that this method significantly outperforms traditional single algorithm models in multiple evaluation metrics. Specifically, in terms of style consistency, the SSIM metric has increased by 12.7%. In terms of composition rationality, F-score has increased by 18.3%; In terms of color harmony, the Δ E * ab value decreased by 23.5%.
Zhang et al. (Thu,) studied this question.