sayurio/anime-art-image

---
license: mit
task_categories:
- image-classification
- image-segmentation
- image-to-text
- image-to-image
- image-text-to-text
- image-to-video
- image-text-to-image
- image-text-to-video
language:
- en
- ja
tags:
- Anime
- Art
- Image
- Manga
- Manhwa
- Cartoon
- Dataset
pretty_name: Anime Art Image
size_categories:
- 10K<n<100K
configs:
- config_name: default
data_files:
- split: train
path: "data/*.zip"
---

# Anime Art Image Archive

## Overview
This repository contains a large-scale, diverse dataset of 2D anime and manga-style artwork. The images have been scraped from multiple web sources to create a comprehensive collection featuring manhwa panels, manga pages, official art, fanart, anime characters, and general 2D illustrations.

## Purpose and Usage
This dataset is published publicly under the MIT license and is intended for **educational, research, and machine learning purposes**. It is exceptionally well-suited for developers, AI researchers, and artists working on:
* **Generative AI:** Training or fine-tuning text-to-image models, LoRAs (Low-Rank Adaptations), and image-to-image pipelines.
* **Computer Vision:** Building image classification, segmentation, and character recognition models.
* **Multimodal Systems:** Training models for image captioning (image-to-text) or video generation from static 2D references.

## Dataset Structure
To optimize download speeds and storage, the dataset has been batched and compressed:
* **File Format:** All images are stored in the highly compressed `WebP` format to maintain visual quality while reducing file size.
* **Organization:** The images are grouped and packaged into `.zip` archives.
* **Batch Size:** Each `.zip` file contains exactly **500 images**.

## Content Warning
Given that this dataset is compiled via automated web scraping of community and art platforms, it may contain a wide variety of artistic expressions. Users should be aware that the dataset is uncurated and may contain content with varying themes, styles, and ratings.

## Copyright and Fair Use Disclaimer
This archive is compiled under the principles of **Fair Use** for computational analysis, machine learning research, and educational study.

* **No Ownership Claimed:** The creator of this repository does not claim any ownership, authorship, or copyright over the original illustrations, characters, or artwork. All rights, title, and interest in the original images remain entirely with their respective illustrators, artists, publishers, and copyright holders.
* **Non-Commercial:** This dataset is provided completely free of charge.
* **Transformative Use:** The aggregation and formatting of these images into computational batches represents a transformative use of publicly accessible web materials for data science.

**Takedown Requests:** If you are an artist, publisher, or copyright holder whose work has been included in this scrape and you wish for it to be removed, please open an issue or contact the repository owner directly. Please provide a reference to the specific ZIP file and the image filename(s) so the content can be located and promptly deleted from the archive.

许可证：MIT协议
任务类别：
- 图像分类
- 图像分割
- 图像到文本
- 图像到图像
- 图像文本到文本
- 图像到视频
- 图像文本到图像
- 图像文本到视频
语言：
- 英语
- 日语
标签：
- 动漫(Anime)
- 美术(Art)
- 图像(Image)
- 漫画(Manga)
- 韩漫(Manhwa)
- 卡通(Cartoon)
- 数据集(Dataset)
美观名称：动漫美术图像
规模类别：10K<n<100K
配置项：
- 配置名称：default
数据文件：
- 拆分：训练集(train)
路径："data/*.zip"

# 动漫美术图像档案库

## 概述
本仓库包含一个大规模、多样化的二维动漫与漫画风格美术作品数据集。我们从多个网络来源抓取图像，构建了涵盖韩漫分镜、漫画页面、官方美术、粉丝创作、动漫角色及通用二维插画的综合藏品。

## 用途与使用
本数据集以MIT协议公开发布，**适用于教育、研究与机器学习场景**。其尤其适配以下开发者、AI研究者及美术创作者的工作：
* **生成式AI**：训练或微调文本到图像模型、LoRA(Low-Rank Adaptation)及图像到图像生成管线。
* **计算机视觉**：构建图像分类、图像分割与字符识别模型。
* **多模态系统**：训练用于图像字幕生成（图像到文本）或基于静态二维参考生成视频的模型。

## 数据集结构
为优化下载速度与存储效率，本数据集采用分批次压缩存储：
* **文件格式**：所有图像均采用高度压缩的WebP格式存储，在维持视觉质量的同时减小文件体积。
* **组织方式**：图像按组打包至.zip归档文件中。
* **单批规模**：每个.zip归档文件恰好包含**500张图像**。

## 内容警示
鉴于本数据集通过自动化抓取社区与美术平台的内容编译而成，其可能包含多样的艺术表达形式。使用者需注意，本数据集未经过人工筛选，可能包含主题、风格与评级各异的内容。

## 版权与合理使用声明
本归档基于**合理使用**原则编译，仅用于计算分析、机器学习研究与教育学习。
* **不主张所有权**：本仓库创建者不对原始插画、角色或美术作品主张任何所有权、署名权或版权。原始图像的全部权利、所有权与利益均归属于其对应的插画师、创作者、出版方及版权持有者。
* **非商业性质**：本数据集完全免费提供。
* **转型性使用**：将这些图像聚合并格式化为计算批次，属于对公开可获取的网络素材进行数据科学层面的转型性使用。

**下架请求**：若您是作品被纳入本次抓取的创作者、出版方或版权持有者，希望移除相关内容，请提交Issue或直接联系仓库所有者。请提供具体的ZIP文件引用与图像文件名，以便我们快速定位并从归档中删除对应内容。