Scholarlink

With the deep integration of artificial intelligence technology into the cultural and creative industries, the generation methods, aesthetic characteristics and interaction mechanisms of digital media art are rapidly evolving. In order to optimize the problems of imprecise style control and insufficient multimodal expression, an art generation system integrating generative adversarial networks, self-attention mechanisms and semantic perception modules is constructed. The results show that the model scores 21.47 on the FID (Fréchet Inception Distance) index, which is significantly better than TransArt and StyleGAN2; and reaches 5.36 on the IS (Inception Score) index, with stronger image expression and style diversity. At the same time, SA-SGAN (Style-Aware Self-Guided Generative Adversarial Network) performs well in multi-dimensional indicators such as style coverage and structure retention rate, with the highest style coverage reaching 0.87. Multiple types of practitioners participated in the actual test evaluation, and the average score of the system was 4.43 points, reflecting strong cross-domain adaptability and user recognition. In summary, this study can expand the technical path of digital art generation driven by AI, and provide a stable and efficient system support solution for intelligent art creation.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on the Innovation Mechanism and Development Strategy of Digital Media Art in the Context of Artificial Intelligence Era

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Tianhui Liang

摘要

With the deep integration of artificial intelligence technology into the cultural and creative industries, the generation methods, aesthetic characteristics and interaction mechanisms of digital media art are rapidly evolving. In order to optimize the problems of imprecise style control and insufficient multimodal expression, an art generation system integrating generative adversarial networks, self-attention mechanisms and semantic perception modules is constructed. The results show that the model scores 21.47 on the FID (Fréchet Inception Distance) index, which is significantly better than TransArt and StyleGAN2; and reaches 5.36 on the IS (Inception Score) index, with stronger image expression and style diversity. At the same time, SA-SGAN (Style-Aware Self-Guided Generative Adversarial Network) performs well in multi-dimensional indicators such as style coverage and structure retention rate, with the highest style coverage reaching 0.87. Multiple types of practitioners participated in the actual test evaluation, and the average score of the system was 4.43 points, reflecting strong cross-domain adaptability and user recognition. In summary, this study can expand the technical path of digital art generation driven by AI, and provide a stable and efficient system support solution for intelligent art creation.