Minthy/BooruCharacters_v0.5

该数据集包含了来自booru画廊的各种角色信息，基于可用数据以及语义和统计分析，确定了每个标签的类别（角色、皮肤、通用角色、别名、伴侣等）、标签之间的关系，并使用了超过2700万张独特图片。对于有足够数量艺术品的角色，数据集提供了流行标签列表和基于图像概括和已知信息的详细文本描述。数据集包含约39,000个独特标签、约33,000个具有评估流行标签的独特角色以及约28,000个独特角色的文本描述。截止日期：2026年3月。

This dataset contains information about a wide variety of characters from booru galleries. Based on the available data, as well as semantic and statistical analysis, the following has been determined: category for each tag (character, skin, generic character, alias, companion, ...), relationships between tags have been identified, and over 27 million unique images were used. For characters with a sufficient number of artworks the dataset provides: list of popular tags, a detailed text description based on image generalization and known information. The dataset contains: ~39,000 unique tags, ~33,000 unique characters with evaluated popular tags, ~28,000 text descriptions for unique characters. Cutoff date: March 2026.