Tag_cleaning: a 1M-10M row Danbooru tag dataset released under Apache-2.0

Grio43 · hf · 2026-09-16

A trending Hugging Face dataset, Tagcleaning by Grio43, contains 1M-10M rows of Danbooru-related tags (likely for cleaning training tags in image generation). It ships in parquet format with pandas/polars/datasets compatibility under an Apache-2.0 license.

Original post →

More from Multimodal

Multimodal channel →