EUVIMEDEUVIMED entdecken
Crossref

Integrating Domain Knowledge into Image Generation Models for Textile Design

Yu Otani, Karol Nowakowski, Masahiro Toyoura

IEEE Computer Graphics and Applications · 2026 · 1-14

DOI: 10.1109/mcg.2026.3709892 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Computer Graphics and Applications, 2026. Verfasst von Yu Otani und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

We propose a prompt augmentation method for textile design image generation that integrates textile-specific domain knowledge, such as patterns, historical context, and design constraints. Recent advances in text-to-image generative models enable designers to create images from natural-language instructions. However, the quality of generated images largely depends on the precision of the prompt, and crafting appropriate prompts remains challenging for designers. To address this issue, our method incorporates domain knowledge about textiles to expand prompts in a way that is suitable for textile design image synthesis. Through qualitative and quantitative evaluations, we confirm that prompts expanded by the proposed method yield better results than those produced from simple prompts or from prompts expanded using only the large language model's internal parametric knowledge.

Zitieren

APA

Otani, Y., Nowakowski, K., & Toyoura, M. (2026). Integrating Domain Knowledge into Image Generation Models for Textile Design. IEEE Computer Graphics and Applications, 1-14. https://doi.org/10.1109/mcg.2026.3709892

Vancouver

Otani Y, Nowakowski K, Toyoura M. Integrating Domain Knowledge into Image Generation Models for Textile Design. IEEE Computer Graphics and Applications. 2026:1-14. doi:10.1109/mcg.2026.3709892

BibTeX
@article{Otani2026,
  author = {Otani, Yu and Nowakowski, Karol and Toyoura, Masahiro},
  title = {Integrating Domain Knowledge into Image Generation Models for Textile Design},
  journal = {IEEE Computer Graphics and Applications},
  year = {2026},
  pages = {1--14},
  doi = {10.1109/mcg.2026.3709892}
}
Kostenlos starten →