Scholarlink

<p>This study proposes an alternative approach to formulating audio description (AD) scripts and delivering them through a purpose-trained artificial intelligence (AI) model capable of automatically generating AD from uploaded photographs of paintings. Building on this multimodal AI system, we developed a mobile application (app), <i>Chromecho</i>, designed to enhance visual arts accessibility for blind and visually impaired (B&amp;VI) visitors to museums and galleries. By simply photographing a painting, either independently or with assistance from companions or museum staff, B&amp;VI users can instantly access AD services for the exhibited artwork. Such an app is particularly valuable in contexts where human-provided AD services are unavailable. Drawing on a review of existing guidelines for visual arts AD, this study proposes a two-component set of guidelines addressing both language and content of painting AD, with the content component centring on genre-specific <i>Artwork-Type Description Guidelines</i>. The rubrics derived from these guidelines serve as benchmarks for training the AI model, ensuring that its outputs align with established best practices in painting AD.</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

Enhancing visual arts accessibility: an AI-driven approach to audio description of paintings

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Xichen Sun,
  • Luyi Zhu,
  • Robin Bautista,
  • Blaithan Altenburg,
  • Qianrui Zhao,
  • Isabella Chan

摘要

This study proposes an alternative approach to formulating audio description (AD) scripts and delivering them through a purpose-trained artificial intelligence (AI) model capable of automatically generating AD from uploaded photographs of paintings. Building on this multimodal AI system, we developed a mobile application (app), Chromecho, designed to enhance visual arts accessibility for blind and visually impaired (B&VI) visitors to museums and galleries. By simply photographing a painting, either independently or with assistance from companions or museum staff, B&VI users can instantly access AD services for the exhibited artwork. Such an app is particularly valuable in contexts where human-provided AD services are unavailable. Drawing on a review of existing guidelines for visual arts AD, this study proposes a two-component set of guidelines addressing both language and content of painting AD, with the content component centring on genre-specific Artwork-Type Description Guidelines. The rubrics derived from these guidelines serve as benchmarks for training the AI model, ensuring that its outputs align with established best practices in painting AD.