README.md

metadata

license: mit
task_categories:
  - image-classification
tags:
  - art
  - image
  - webdataset
  - datasets
size_categories:
  - 100K<n<1M

Anime Image Classification Dataset (v1.5)

This is the webdataset dataset, containing 323060 images in total.

Images here are resized to min(width, height) <= 640.

How to Use It

from datasets import load_dataset

dataset = load_dataset('just-a-try/anime-classification-v1.5')
print(dataset["train"][0])

Images

323060 images in total.

Split	Image Count	Total Size
train	257996	14.5 GB
test	32506	1.83 GB
val	32558	1.84 GB

Class	Image Count	Total Size
3d	48200	1.98 GB
bangumi	40815	1.18 GB
comic	60147	5.21 GB
illustration	122344	6.66 GB
not_painting	51554	3.15 GB

Citation

@inproceedings{anime-cls-v1_5,
  author = {narugo1992},
  title = {Anime Image Classification Dataset (v1.5)},
  year = {2025},
  url = {https://huggingface.co/narugo1992},
  note = {Dataset available at https://huggingface.co/datasets/just-a-try/anime-classification-v1.5}
}