CyberHarem/kikyou_bluearchive

---
license: mit
task_categories:
- text-to-image
tags:
- art
- not-for-all-audiences
size_categories:
- n<1K
---

# Dataset of kikyou/桐生キキョウ/桔梗 (Blue Archive)

This is the dataset of kikyou/桐生キキョウ/桔梗 (Blue Archive), containing 500 images and their tags.

The core tags of this character are `animal_ears, black_hair, cat_ears, short_hair, halo, black_eyes, blue_halo, tail, cat_tail, multiple_tails, two_tails`, which are pruned in this dataset.

Images are crawled from many sites (e.g. danbooru, pixiv, zerochan ...), the auto-crawling system is powered by [DeepGHS Team](https://github.com/deepghs)([huggingface organization](https://huggingface.co/deepghs)).

## List of Packages

| Name | Images | Size | Download | Type | Description |
|:-----------------|---------:|:-----------|:--------------------------------------------------------------------------------------------------------------------|:-----------|:---------------------------------------------------------------------|
| raw | 500 | 877.49 MiB | [Download](https://huggingface.co/datasets/CyberHarem/kikyou_bluearchive/resolve/main/dataset-raw.zip) | Waifuc-Raw | Raw data with meta information (min edge aligned to 1400 if larger). |
| 1200 | 500 | 721.47 MiB | [Download](https://huggingface.co/datasets/CyberHarem/kikyou_bluearchive/resolve/main/dataset-1200.zip) | IMG+TXT | dataset with the shorter side not exceeding 1200 pixels. |
| stage3-p480-1200 | 1304 | 1.44 GiB | [Download](https://huggingface.co/datasets/CyberHarem/kikyou_bluearchive/resolve/main/dataset-stage3-p480-1200.zip) | IMG+TXT | 3-stage cropped dataset with the area not less than 480x480 pixels. |

### Load Raw Dataset with Waifuc

We provide raw dataset (including tagged images) for [waifuc](https://deepghs.github.io/waifuc/main/tutorials/installation/index.html) loading. If you need this, just run the following code

```python
import os
import zipfile

from huggingface_hub import hf_hub_download
from waifuc.source import LocalSource

# download raw archive file
zip_file = hf_hub_download(
repo_id='CyberHarem/kikyou_bluearchive',
repo_type='dataset',
filename='dataset-raw.zip',
)

# extract files to your directory
dataset_dir = 'dataset_dir'
os.makedirs(dataset_dir, exist_ok=True)
with zipfile.ZipFile(zip_file, 'r') as zf:
zf.extractall(dataset_dir)

# load the dataset with waifuc
source = LocalSource(dataset_dir)
for item in source:
print(item.image, item.meta['filename'], item.meta['tags'])
```

## List of Clusters

List of tag clustering result, maybe some outfits can be mined here.

### Raw Text Version

| # | Samples | Img-1 | Img-2 | Img-3 | Img-4 | Img-5 | Tags |
|----:|----------:|:--------------------------------|:--------------------------------|:--------------------------------|:--------------------------------|:--------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
| 0 | 9 |  |  |  |  |  | 1girl, completely_nude, blush, nipples, 1boy, medium_breasts, collarbone, hetero, solo_focus, looking_at_viewer, sweat, heart, navel, open_mouth, pubic_hair, sex, simple_background |
| 1 | 7 |  |  |  |  |  | 1boy, 1girl, black_choker, blush, from_side, hetero, solo_focus, white_shirt, erection, sailor_collar, saliva, uncensored, animal_ear_fluff, school_uniform, veiny_penis, testicles, blue_neckerchief, closed_eyes, crying, cum_in_mouth, huge_penis, irrumatio, large_penis, nude, tears |
| 2 | 7 |  |  |  |  |  | 1boy, 1girl, blue_neckerchief, blush, hetero, long_sleeves, serafuku, white_shirt, black_nails, black_sailor_collar, censored, erection, nail_polish, solo_focus, fingernails, ribbon_choker, black_choker, animal_ear_fluff, breasts, closed_mouth, cum, handjob, looking_at_penis, simple_background |
| 3 | 25 |  |  |  |  |  | 1girl, black_sailor_collar, black_skirt, blue_neckerchief, fingernails, long_sleeves, pleated_skirt, serafuku, solo, nail_polish, white_background, black_nails, looking_at_viewer, simple_background, haori, closed_mouth, blush |
| 4 | 5 |  |  |  |  |  | 1girl, black_nails, black_sailor_collar, blue_neckerchief, haori, long_sleeves, looking_at_viewer, nail_polish, serafuku, solo, upper_body, choker, closed_mouth, fingernails, makeup |
| 5 | 5 |  |  |  |  |  | 1girl, black_sailor_collar, blue_neckerchief, choker, haori, looking_at_viewer, serafuku, simple_background, solo, upper_body, white_background, closed_mouth, collarbone, cropped_torso, long_sleeves |
| 6 | 11 |  |  |  |  |  | 1girl, barefoot, black_nails, black_sailor_collar, blue_neckerchief, long_sleeves, looking_at_viewer, pleated_skirt, serafuku, solo, toenail_polish, toes, black_skirt, closed_mouth, indoors, sitting, fingernails, white_shirt, holding_book, soles, bare_legs, choker, foot_focus, knees_up |
| 7 | 13 |  |  |  |  |  | 1girl, looking_at_viewer, solo, closed_mouth, collarbone, navel, alternate_costume, medium_breasts, simple_background, blush, cleavage, white_background, black_bikini, black_nails, nail_polish, stomach, cowboy_shot, fingernails, side-tie_bikini_bottom |

### Table Version

| # | Samples | Img-1 | Img-2 | Img-3 | Img-4 | Img-5 | 1girl | completely_nude | blush | nipples | 1boy | medium_breasts | collarbone | hetero | solo_focus | looking_at_viewer | sweat | heart | navel | open_mouth | pubic_hair | sex | simple_background | black_choker | from_side | white_shirt | erection | sailor_collar | saliva | uncensored | animal_ear_fluff | school_uniform | veiny_penis | testicles | blue_neckerchief | closed_eyes | crying | cum_in_mouth | huge_penis | irrumatio | large_penis | nude | tears | long_sleeves | serafuku | black_nails | black_sailor_collar | censored | nail_polish | fingernails | ribbon_choker | breasts | closed_mouth | cum | handjob | looking_at_penis | black_skirt | pleated_skirt | solo | white_background | haori | upper_body | choker | makeup | cropped_torso | barefoot | toenail_polish | toes | indoors | sitting | holding_book | soles | bare_legs | foot_focus | knees_up | alternate_costume | cleavage | black_bikini | stomach | cowboy_shot | side-tie_bikini_bottom |
|----:|----------:|:--------------------------------|:--------------------------------|:--------------------------------|:--------------------------------|:--------------------------------|:--------|:------------------|:--------|:----------|:-------|:-----------------|:-------------|:---------|:-------------|:--------------------|:--------|:--------|:--------|:-------------|:-------------|:------|:--------------------|:---------------|:------------|:--------------|:-----------|:----------------|:---------|:-------------|:-------------------|:-----------------|:--------------|:------------|:-------------------|:--------------|:---------|:---------------|:-------------|:------------|:--------------|:-------|:--------|:---------------|:-----------|:--------------|:----------------------|:-----------|:--------------|:--------------|:----------------|:----------|:---------------|:------|:----------|:-------------------|:--------------|:----------------|:-------|:-------------------|:--------|:-------------|:---------|:---------|:----------------|:-----------|:-----------------|:-------|:----------|:----------|:---------------|:--------|:------------|:-------------|:-----------|:--------------------|:-----------|:---------------|:----------|:--------------|:-------------------------|
| 0 | 9 |  |  |  |  |  | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| 1 | 7 |  |  |  |  |  | X | | X | | X | | | X | X | | | | | | | | | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | X | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
| 2 | 7 |  |  |  |  |  | X | | X | | X | | | X | X | | | | | | | | X | X | | X | X | | | | X | | | | X | | | | | | | | | X | X | X | X | X | X | X | X | X | X | X | X | X | | | | | | | | | | | | | | | | | | | | | | | | | |
| 3 | 25 |  |  |  |  |  | X | | X | | | | | | | X | | | | | | | X | | | | | | | | | | | | X | | | | | | | | | X | X | X | X | | X | X | | | X | | | | X | X | X | X | X | | | | | | | | | | | | | | | | | | | | |
| 4 | 5 |  |  |  |  |  | X | | | | | | | | | X | | | | | | | | | | | | | | | | | | | X | | | | | | | | | X | X | X | X | | X | X | | | X | | | | | | X | | X | X | X | X | | | | | | | | | | | | | | | | | |
| 5 | 5 |  |  |  |  |  | X | | | | | | X | | | X | | | | | | | X | | | | | | | | | | | | X | | | | | | | | | X | X | | X | | | | | | X | | | | | | X | X | X | X | X | | X | | | | | | | | | | | | | | | | |
| 6 | 11 |  |  |  |  |  | X | | | | | | | | | X | | | | | | | | | | X | | | | | | | | | X | | | | | | | | | X | X | X | X | | | X | | | X | | | | X | X | X | | | | X | | | X | X | X | X | X | X | X | X | X | X | | | | | | |
| 7 | 13 |  |  |  |  |  | X | | X | | | X | X | | | X | | | X | | | | X | | | | | | | | | | | | | | | | | | | | | | | X | | | X | X | | | X | | | | | | X | X | | | | | | | | | | | | | | | | X | X | X | X | X | X |

原始信息汇总

数据集概述

数据集基本信息

• 名称: Dataset of kikyou/桐生キキョウ/桔梗 (Blue Archive)
• 许可证: MIT
• 任务类别: text-to-image
• 标签: art, not-for-all-audiences
• 大小类别: n<1K

数据集内容

• 图像数量: 500
• 核心标签: animal_ears, black_hair, cat_ears, short_hair, halo, black_eyes, blue_halo, tail, cat_tail, multiple_tails, two_tails

数据包列表

名称	图像数量	大小	类型	描述
raw	500	877.49 MiB	Waifuc-Raw	包含元信息的原始数据（最小边对齐到1400像素，如果更大）
1200	500	721.47 MiB	IMG+TXT	短边不超过1200像素的数据集
stage3-p480-1200	1304	1.44 GiB	IMG+TXT	3阶段裁剪数据集，区域不小于480x480像素

数据集加载

• 加载工具: waifuc

• 加载代码: python import os import zipfile from huggingface_hub import hf_hub_download from waifuc.source import LocalSource

  下载原始归档文件

  zip_file = hf_hub_download( repo_id=CyberHarem/kikyou_bluearchive, repo_type=dataset, filename=dataset-raw.zip, )

  提取文件到指定目录

  dataset_dir = dataset_dir os.makedirs(dataset_dir, exist_ok=True) with zipfile.ZipFile(zip_file, r) as zf: zf.extractall(dataset_dir)

  使用waifuc加载数据集

  source = LocalSource(dataset_dir) for item in source: print(item.image, item.meta[filename], item.meta[tags])

标签聚类结果

原始文本版本

#	样本数量	图像1	图像2	图像3	图像4	图像5	标签
0	9						1girl, completely_nude, blush, nipples, 1boy, medium_breasts, collarbone, hetero, solo_focus, looking_at_viewer, sweat, heart, navel, open_mouth, pubic_hair, sex, simple_background
1	7						1boy, 1girl, black_choker, blush, from_side, hetero, solo_focus, white_shirt, erection, sailor_collar, saliva, uncensored, animal_ear_fluff, school_uniform, veiny_penis, testicles, blue_neckerchief, closed_eyes, crying, cum_in_mouth, huge_penis, irrumatio, large_penis, nude, tears
2	7						1boy, 1girl, blue_neckerchief, blush, hetero, long_sleeves, serafuku, white_shirt, black_nails, black_sailor_collar, censored, erection, nail_polish, solo_focus, fingernails, ribbon_choker, black_choker, animal_ear_fluff, breasts, closed_mouth, cum, handjob, looking_at_penis, simple_background
3	25						1girl, black_sailor_collar, black_skirt, blue_neckerchief, fingernails, long_sleeves, pleated_skirt, serafuku, solo, nail_polish, white_background, black_nails, looking_at_viewer, simple_background, haori, closed_mouth, blush
4	5						1girl, black_nails, black_sailor_collar, blue_neckerchief, haori, long_sleeves, looking_at_viewer, nail_polish, serafuku, solo, upper_body, choker, closed_mouth, fingernails, makeup
5	5						1girl, black_sailor_collar, blue_neckerchief, choker, haori, looking_at_viewer, serafuku, simple_background, solo, upper_body, white_background, closed_mouth, collarbone, cropped_torso, long_sleeves
6	11						1girl, barefoot, black_nails, black_sailor_collar, blue_neckerchief, long_sleeves, looking_at_viewer, pleated_skirt, serafuku, solo, toenail_polish, toes, black_skirt, closed_mouth, indoors, sitting, fingernails, white_shirt, holding_book, soles, bare_legs, choker, foot_focus, knees_up
7	13						1girl, looking_at_viewer, solo, closed_mouth, collarbone, navel, alternate_costume, medium_breasts, simple_background, blush, cleavage, white_background, black_bikini, black_nails, nail_polish, stomach, cowboy_shot, fingernails, side-tie_bikini_bottom

搜集汇总

数据集介绍

构建方式

在二次元图像生成领域，高质量的角色数据集是模型微调与风格迁移的重要基石。本数据集聚焦于《蔚蓝档案》中的人气角色桐生桔梗，通过自动化爬虫系统从Danbooru、Pixiv、Zerochan等多个主流图站采集原始图像，并由DeepGHS团队提供技术支撑。数据集共收录500张图像，每张图像均附带详尽的标签元信息。核心角色特征标签如兽耳、黑发、猫耳、短发、光环、黑瞳、蓝色光环、尾巴、猫尾、多尾及双尾等已被系统性地剪枝处理，以突出角色辨识度。数据集提供了三种封装形式：原始数据包保留完整元数据并统一将最小边对齐至1400像素；1200版本将短边限制在1200像素以内；而stage3-p480-1200版本则采用三级裁剪策略，确保每个子区域不小于480×480像素。

特点

该数据集最显著的特点在于其多维度的结构化设计。原始数据包支持通过Waifuc框架直接加载，用户可便捷地获取图像、文件名及标签信息，为二次创作提供灵活的数据接口。经过尺寸归一化处理的1200版本，在保持图像质量的同时显著降低了存储开销，适合快速批量处理。三级裁剪版本通过智能分区策略，将单张图像扩展为多个局部区域，总计生成1304个样本，极大丰富了训练数据的多样性。此外，数据集内置了标签聚类分析结果，以可视化方式呈现了包括校服、泳装、日常装束在内的多种服饰搭配，以及不同场景下的角色互动模式，为下游任务如风格迁移、角色换装等提供了清晰的语义线索。

使用方法

使用本数据集时，研究者可根据任务需求选择不同的数据包。若需保留完整的元数据与高分辨率图像，推荐下载原始数据包，并通过Waifuc的LocalSource模块进行加载，示例代码展示了从HuggingFace Hub下载压缩包、解压至本地目录以及迭代访问图像与标签的完整流程。对于追求训练效率的场景，1200版本或三级裁剪版本更为合适，二者均以图像-标签配对格式提供，可直接接入常见的文本到图像生成框架。聚类结果以表格和图像样本双重视角呈现，便于用户快速筛选特定场景或服饰的子集，从而定制更具针对性的训练数据。整体而言，该数据集兼顾了原始数据的完整性与处理后的易用性，适配从学术研究到工业落地的多种应用场景。

背景与挑战

背景概述

在文本到图像生成领域，尤其是面向动漫角色创作时，高质量、带标注的专用数据集是模型训练的关键基石。CyberHarem/kikyou_bluearchive 数据集由 DeepGHS 团队于近期创建，聚焦于游戏《蔚蓝档案》中的角色“桐生桔梗”。该数据集包含 500 张经过精心标注的图像，核心特征如猫耳、黑发、光环等标签被系统化整理，旨在为动漫风格的角色生成模型提供精准的视觉与语义对齐资源。作为开源项目，它依托 waifuc 框架实现了数据的高效爬取与标准化处理，为角色定制化生成研究提供了可复现的范例，推动了二次元图像生成领域从通用模型向特定角色精细化控制的演进。

当前挑战

数据集面临多重挑战。首先，在领域问题层面，动漫角色生成需解决风格一致性、姿态多样性及复杂标签（如“异色瞳”、“多重尾巴”）的语义精准映射，现有模型易产生特征混淆或细节缺失。其次，构建过程中，数据爬取面临版权合规与来源异构性（如 danbooru 标签体系与 pixiv 元数据不统一），导致标注噪声累积。此外，图像分辨率差异（从 480x480 到 1200 像素）及内容敏感标签（如 NSFW 内容）的过滤与平衡，增加了数据清洗与隐私保护的复杂度，限制了数据集的泛化能力与伦理合规性。

常用场景

经典使用场景

在二次元图像生成领域，CyberHarem/kikyou_bluearchive 数据集为基于扩散模型的文本到图像任务提供了高质量的素材。该数据集包含500张《碧蓝档案》角色桐生桔梗的图像及其标签，聚焦于猫耳、黑发、光环等核心视觉特征。研究人员常将其用于微调Stable Diffusion等预训练模型，以生成风格一致的角色形象，或作为少样本学习的基准数据集，探索在有限样本下如何保持角色身份特征的稳定性。

解决学术问题

该数据集有效解决了动漫角色个性化生成中的数据稀缺与标注不一致问题。通过提供经裁剪和标准化处理的图像-文本对，它支持了角色概念解耦研究，即如何从多模态数据中分离出角色身份（如桔梗的猫耳、双尾）与背景、姿势等无关属性。此外，其标签聚类结果（如制服、泳装等变体）为细粒度视觉属性识别和组合生成提供了实验基础，推动了可控图像生成领域的发展。

衍生相关工作

该数据集衍生了一系列经典工作，包括基于LoRA（Low-Rank Adaptation）的高效角色微调方法，以及利用其标签聚类结果进行多概念融合生成的探索。此外，研究者借助该数据集的原始元数据，开发了针对动漫图像的自动标注与清洗算法（如DeepGHS团队的爬取系统）。这些工作不仅提升了角色生成的质量，还促进了开源社区中动漫数据集标准化流程的建立。