NEXTAltair/genai-image-tag-db-mit

该数据集是一个用于生成人工智能图像的标签数据库，主要应用于稳定扩散（Stable Diffusion）和LoRA等技术。数据集包含来自多个来源（如Danbooru、e621和Derpibooru）的标签数据，提供了SQLite和Parquet两种格式的数据文件。SQLite数据库包含完整的关联模式，而Parquet文件则专门用于Hugging Face数据集查看器（目前仅包含Danbooru数据）。数据集还包含标签的状态信息（如是否弃用、弃用时间等）以及多语言标签（如日语、中文等）。数据集的许可证为MIT，并包含了来自多个MIT许可来源的数据。数据清理过程确保了多语言标签的质量，例如仅保留包含特定字符集的翻译。

This dataset is a tag database for generative AI images, primarily used in technologies like Stable Diffusion and LoRA. It includes tag data from multiple sources such as Danbooru, e621, and Derpibooru, and provides data files in both SQLite and Parquet formats. The SQLite database features a full relational schema, while the Parquet files are specifically for the Hugging Face Dataset Viewer (currently Danbooru only). The dataset also includes tag status information (e.g., deprecated flags, deprecation time) and multilingual tags (e.g., Japanese, Chinese). The dataset is licensed under MIT and incorporates data from several MIT-licensed sources. Data cleanup processes ensure the quality of multilingual tags, such as retaining only translations containing specific character sets.