| Abstract |
| This study proposes a novel approach to visually restoring Goguryeo-period clothing using generative artificial intelligence (AI). To achieve this, visual characteristics of Goguryeo attire were extracted from sources considered closely related with the period, including Goguryeo tomb murals, haniwa (clay figurines), Saekdong (traditional multicolored stripes), and ancient textiles excavated across Asia. These elements informed the visual reconstruction of Goguryeo clothing through generative AI. First, clothing characteristics from the Three Kingdoms period were compiled through a literature review. Reference images for restoration were then selected, and tailored prompts were designed for DALL-E to generate restored images. The generated images were analyzed to determine how Goguryeo clothing was visually represented. The visual completeness and historical appropriateness of the restored images, which reflected features found in tomb murals and haniwa, were also assessed. Specifically, by incorporating Saekdong and traditional pattern-dyed textiles unearthed from various regions of North and East Asia, the study explored possible applications of these fabrics in ancient attire. Generative AI played a meaningful role in the visual reconstruction of traditional garments. This study is significant in that it demonstrates the potential of generative artificial intelligence for visually restoring ancient clothing and thereby offers a new solution to the visual limitations traditionally encountered in historical costume research. |
|
|
| Key Words |
| Generative AI, Clothing Restoration, Goguryeo Tomb Murals, Haniwa, Saekdong. Geum & Gye, 생성형 인공지능, 복식 복원, 고구려 고분벽화, 하니와, 색동, 금계 직물 |
|
|
 |
|
