Scholarlink

<p>The convergence of artificial intelligence and digital media art is transforming contemporary creative practices by enabling adaptive, participatory, and context-aware artistic systems. Recent advances in representation learning and multimodal intelligence have expanded the role of AI from a passive content generator to an active creative collaborator. This study introduces an Adaptive Aesthetic intelligence (AAI) framework that integrates variational autoencoders (VAEs) with attention-based neural style modulation to generate interactive digital artworks that evolve in response to real-time user interaction and contextual signals. The framework is trained and evaluated using the AI-Artwork Dataset, a large-scale Kaggle collection comprising both AI-generated and human-created art images, enabling systematic comparative analysis of aesthetic patterns, stylistic diversity, and perceptual realism. Latent-space compression via autoencoder-based dimensionality reduction ensures computational efficiency and smooth real-time responsiveness. Quantitative and user-centric evaluations demonstrate strong performance, achieving aesthetic realism of 96.8%. Compared with conventional neural art generation approaches, the proposed framework improves adaptive responsiveness by 18.7% while preserving higher visual coherence across interactive transitions. These results validate the effectiveness of adaptive AI-driven co-creative systems and highlight their potential to extend digital media art beyond static, predefined creative paradigms.</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

Computational co-creation in digital media art: adaptive aesthetic intelligence for interactive visual expression

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月3日

查看全文
  • Yu Liu,
  • Haozhe Bai,
  • Yunlong Li

摘要

The convergence of artificial intelligence and digital media art is transforming contemporary creative practices by enabling adaptive, participatory, and context-aware artistic systems. Recent advances in representation learning and multimodal intelligence have expanded the role of AI from a passive content generator to an active creative collaborator. This study introduces an Adaptive Aesthetic intelligence (AAI) framework that integrates variational autoencoders (VAEs) with attention-based neural style modulation to generate interactive digital artworks that evolve in response to real-time user interaction and contextual signals. The framework is trained and evaluated using the AI-Artwork Dataset, a large-scale Kaggle collection comprising both AI-generated and human-created art images, enabling systematic comparative analysis of aesthetic patterns, stylistic diversity, and perceptual realism. Latent-space compression via autoencoder-based dimensionality reduction ensures computational efficiency and smooth real-time responsiveness. Quantitative and user-centric evaluations demonstrate strong performance, achieving aesthetic realism of 96.8%. Compared with conventional neural art generation approaches, the proposed framework improves adaptive responsiveness by 18.7% while preserving higher visual coherence across interactive transitions. These results validate the effectiveness of adaptive AI-driven co-creative systems and highlight their potential to extend digital media art beyond static, predefined creative paradigms.