Scholarlink

<p>This paper introduces an intelligent art design system powered by generative AI to address the cross-modal semantic disconnection in art creation. By integrating symbolic cognitive constraints and dynamic rule injection mechanisms, the system enhances both semantic accuracy and aesthetic quality in generative models. A cultural semantic knowledge graph is developed to map cultural metaphors to visual representations, while a symbolic attention encoder ensures semantic consistency during image generation. Additionally, a differentiable aesthetic rule validator adjusts the output in real-time to optimize the alignment with both artistic intent and aesthetic standards. Experimental results show that the proposed system outperforms existing models in key areas such as cultural symbol mapping (92.5% coverage), symmetry, and color harmony, delivering more accurate and aesthetically pleasing results. The system has been applied in real-world art creation, significantly improving the efficiency and creativity of artists by providing culturally relevant and visually coherent references. The findings highlight the system’s potential as an innovative tool for generative art, bridging the gap between semantic accuracy and aesthetic quality in the creative process.</p>

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 construction of intelligent art design system based on multimodal perception and generative AI

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月19日

查看全文
  • Hang Hang,
  • Zhengdong Li

摘要

This paper introduces an intelligent art design system powered by generative AI to address the cross-modal semantic disconnection in art creation. By integrating symbolic cognitive constraints and dynamic rule injection mechanisms, the system enhances both semantic accuracy and aesthetic quality in generative models. A cultural semantic knowledge graph is developed to map cultural metaphors to visual representations, while a symbolic attention encoder ensures semantic consistency during image generation. Additionally, a differentiable aesthetic rule validator adjusts the output in real-time to optimize the alignment with both artistic intent and aesthetic standards. Experimental results show that the proposed system outperforms existing models in key areas such as cultural symbol mapping (92.5% coverage), symmetry, and color harmony, delivering more accurate and aesthetically pleasing results. The system has been applied in real-world art creation, significantly improving the efficiency and creativity of artists by providing culturally relevant and visually coherent references. The findings highlight the system’s potential as an innovative tool for generative art, bridging the gap between semantic accuracy and aesthetic quality in the creative process.