Oracle Bone Character Generation with Diffusion Models

• Item type: SIG Technical Reports(1)
• 公開日: 2024-06-22
• タイトル: Oracle Bone Character Generation with Diffusion Models
• タイトル / 言語: en
• タイトル: Oracle Bone Character Generation with Diffusion Models
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: CG一般セッション
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_18gh
• 資源タイプ: technical report
• 著者所属: Japan Advanced Institute of Science and Technology
• 著者所属: Japan Advanced Institute of Science and Technology
• 著者所属: Japan Advanced Institute of Science and Technology
• 著者所属: Japan Advanced Institute of Science and Technology
• 著者所属(英): en
• 著者所属(英): Japan Advanced Institute of Science and Technology
• 著者所属(英): en
• 著者所属(英): Japan Advanced Institute of Science and Technology
• 著者所属(英): en
• 著者所属(英): Japan Advanced Institute of Science and Technology
• 著者所属(英): en
• 著者所属(英): Japan Advanced Institute of Science and Technology
• 著者名: Xiaoxuan, Xie; Xusheng, Du; Minhao, Li; Haoran, Xie
• 著者名(英): Xiaoxuan, Xie; Xusheng, Du; Minhao, Li; Haoran, Xie
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: Generating is a challenging task due to their scarcity and unique pictographic nature. In this research, we propose a novel approach to generate oracle bone character-style images corresponding to the given image input using diffusion models. Unlike existing datasets of oracle bone characters, they typically store paired characters with modern Chinese characters. We first construct the dataset that aligns oracle bone characters, text prompts, and object images. We then train a stable diffusion model on this dataset. By inputting images of different objects combined with corresponding text prompts, the model generates the corresponding images in the style of oracle bone characters. Additionally, we integrate an optimization module to refine the initial results, ensuring they better conform to the original structure and norms of oracle bone characters. The qualitative evaluation demonstrates that our model excels in generating oracle bone character-style images that are both stylistically and semantically consistent.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: Generating is a challenging task due to their scarcity and unique pictographic nature. In this research, we propose a novel approach to generate oracle bone character-style images corresponding to the given image input using diffusion models. Unlike existing datasets of oracle bone characters, they typically store paired characters with modern Chinese characters. We first construct the dataset that aligns oracle bone characters, text prompts, and object images. We then train a stable diffusion model on this dataset. By inputting images of different objects combined with corresponding text prompts, the model generates the corresponding images in the style of oracle bone characters. Additionally, we integrate an optimization module to refine the initial results, ensuring they better conform to the original structure and norms of oracle bone characters. The qualitative evaluation demonstrates that our model excels in generating oracle bone character-style images that are both stylistically and semantically consistent.
• 書誌レコードID / 収録物識別子タイプ: NCID
• 書誌レコードID / 収録物識別子: AN10100541
• 書誌情報: 研究報告コンピュータグラフィックスとビジュアル情報学（CG）; 巻 2024-CG-194, 号 4, p. 1-6, 発行日 2024-06-22
• ISSN / 収録物識別子タイプ: ISSN
• ISSN / 収録物識別子: 2188-8949
• Notice: SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc.
• 出版者 / 言語: ja
• 出版者: 情報処理学会