PixelPerv/danbooru-tags-20260518-duckdb

该数据集是原始Danbooru标签数据集（u-haru/danbooru-tags-20260518）的优化和精简衍生版本，专门设计用于促进面向插图和动漫AI模型的标签分析。它通过仅保留分析工作负载必需的列来优化大小，显著减少文件大小和下载时间；采用严格的数据类型转换（如BIGINT、TIMESTAMP WITH TIME ZONE和结构化数组VARCHAR[]）以加速查询处理；并将所有数据打包成单个DuckDB数据库文件，便于本地探索和WebUI下载。数据集包含Danbooru帖子的详细信息，如唯一标识符、创建时间、上传者ID、评分、投票数、收藏数、内容评级（g、s、q、e）以及多种标签数组（如通用标签、角色标签、版权标签、艺术家标签和元标签）。This dataset is an optimized, streamlined derivative of the original Danbooru tags dataset (u-haru/danbooru-tags-20260518), specifically engineered to facilitate tag analysis for Illustrious and Anime-focused AI models. It features size optimization by retaining only essential columns for analytical workloads, drastically reducing file size and download times; strict data typing (e.g., BIGINT, TIMESTAMP WITH TIME ZONE, and structured arrays VARCHAR[]) to accelerate query processing; and packaging all data into a single DuckDB database file for easy local exploration and webUI downloading. The dataset includes detailed information on Danbooru posts, such as unique identifiers, creation timestamps, uploader IDs, scores, vote counts, favorite counts, content ratings (g, s, q, e), and various tag arrays (e.g., general tags, character tags, copyright tags, artist tags, and meta tags).