Skip to content

Dataset · animelover

Danbooru images for text-to-image

animelover/danbooru2022

Around 4 million anime images from Danbooru for training text-to-image models.

What's inside

Images gathered from Danbooru, totaling approximately 4 million or more. The card specifies a post ID range of 6019085 to 1019085. All images with a short side exceeding 768 pixels were resized to 768. Some images are missing due to requiring a gold account, the exclusion of images over 25 million pixels, and filtered tags (furry, realistic, 3d, retro styles, scans, etc.).

What you can do with it

  • Training anime-style text-to-image models
  • Fine-tuning LoRA for Stable Diffusion on Danbooru
  • Validating anime image generation
Tasks
Text → image
Languages
English
Size
1M–10M rows
Downloads this month
1.8K
Likes
167

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("animelover/danbooru2022")