Dataset · deepghs
8 million anime illustrations with Danbooru tags
deepghs/danbooru2024
A massive archive of 8.05 million anime art pieces with user tags and descriptions for training generative models.
What's inside
The dataset includes around 8.05 million original illustrations and anime artworks from the Danbooru platform. Each image is accompanied by detailed user tags and text descriptions in English and Japanese, with metadata packaged in parquet format.
What you can do with it
- Training and fine-tuning anime and illustration text-to-image models
- Training multimodal autotaggers for image labeling
- Training zero-shot art classification systems
- Tasks
- Image classification, Zero shot image classification, Text → image
- Languages
- English, Japanese
- Size
- 1M–10M rows
- Downloads this month
- 4K
- Likes
- 66
- Access
- on request on Hugging Face
Models trained on it
- Claquasse/foot_anime_yoloObject detection 914
- John6666/seele-noobai-sdxl-v21-sdxlText → image 6
- igidn/DT24-TinyImage classification 6
- John6666/illumiyume-xl-illustrious-v32-vpred-sdxlText → image 5
- John6666/nuke-ilyume-tuned-nsfw-v20-sdxlText → image 4
- John6666/illumiyume-xl-illustrious-v20-sdxlText → image 3
- John6666/illumiyume-xl-illustrious-v30-sdxlText → image 3
- John6666/illumiyume-xl-illustrious-v31-sdxlText → image 3
How to load it
from datasets import load_dataset
ds = load_dataset("deepghs/danbooru2024")