Team Ai
Datasetpublic

cloud19/gelbooru-characters-enriched

Gelbooru Characters Enriched This dataset is an enriched, fully-mapped version of Gelbooru character tags. It contains resolved franchise (copyright) associations and core appearance features (core tags) for 263,441 unique characters. Dataset Details The dataset maps the original character list to their corresponding copyrights (franchises) and general core attributes. It was constructed using a multi-stage hybrid extraction pipeline: Regex Extraction: Extracting… See the full description on the dataset page: https://huggingface.co/datasets/cloud19/gelbooru-characters-enriched.

sourceHugging Facecc-by-4.0updated 3mo agoView on Hugging Face
0likes27downloads
Dataset Card

Gelbooru Characters Enriched

This dataset is an enriched, fully-mapped version of Gelbooru character tags. It contains resolved franchise (copyright) associations and core appearance features (core tags) for 263,441 unique characters.

Dataset Details

The dataset maps the original character list to their corresponding copyrights (franchises) and general core attributes. It was constructed using a multi-stage hybrid extraction pipeline:

  1. 1.Regex Extraction: Extracting copyrights from name parentheses (e.g., _(genshin_impact)).
  2. 2.Gelbooru DAPI Resolution: Fetching posts and querying tag types directly from Gelbooru to map copyrights and general tags.
  3. 3.Danbooru Fallback API: Fetching metadata from Danbooru posts, with automated tag alias resolution.

Schema

The dataset is saved in CSV format and contains the following columns:

Column NameTypeDescription
tag_idintegerUnique identifier for the character tag on Gelbooru
tag_namestringName of the character tag
post_countintegerTotal number of posts on Gelbooru tagged with this character
copyrightstringResolved franchise or copyright associated with the character
core_tagsstringComma-separated list of the most frequent general tags representing the character's appearance

Verification and Statistics

  • —Total Records: 263,441
  • —Populated Copyrights: 246,856 (93.7% coverage)
  • —Populated Core Tags: 235,822 (89.5% coverage)