CyberHarem/ichigaya_arisa_bangdream
收藏Hugging Face2024-01-15 更新2024-03-04 收录
下载链接:
https://hf-mirror.com/datasets/CyberHarem/ichigaya_arisa_bangdream
下载链接
链接失效反馈官方服务:
资源简介:
这是一个关于市ヶ谷有咲(BanG Dream!)角色的数据集,包含500张图片及其标签。图片来源于多个网站(如danbooru、pixiv、zerochan等),并由DeepGHS团队的自动爬取系统生成。该数据集的核心标签包括`blonde_hair, long_hair, bangs, hair_ornament, twintails, x_hair_ornament, brown_eyes, sidelocks, breasts, yellow_eyes`。数据集提供了不同尺寸和格式的下载链接,并详细说明了每个数据包的内容和用途。
This is a dataset centered on the character Arisa Ichigaya from *BanG Dream!*. It contains 500 images along with their corresponding labels. Images are sourced from multiple online platforms including danbooru, pixiv, zerochan, etc., and were collected via the automatic crawling system developed by the DeepGHS team. The core annotated tags of this dataset include `blonde_hair, long_hair, bangs, hair_ornament, twintails, x_hair_ornament, brown_eyes, sidelocks, breasts, yellow_eyes`. The dataset provides download links for various sizes and formats, and elaborates on the content and intended use of each data package.
提供机构:
CyberHarem
原始信息汇总
数据集概述
基本信息
- 数据集名称: Dataset of ichigaya_arisa/市ヶ谷有咲 (BanG Dream!)
- 许可证: MIT
- 任务类别: text-to-image
- 标签: art, not-for-all-audiences
- 大小类别: n<1K
数据内容
- 图像数量: 500
- 核心标签:
blonde_hair, long_hair, bangs, hair_ornament, twintails, x_hair_ornament, brown_eyes, sidelocks, breasts, yellow_eyes
数据包列表
| 名称 | 图像数量 | 大小 | 类型 | 描述 |
|---|---|---|---|---|
| raw | 500 | 699.20 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)。 |
| 800 | 500 | 381.84 MiB | IMG+TXT | 短边不超过800像素的数据集。 |
| stage3-p480-800 | 1257 | 857.18 MiB | IMG+TXT | 3阶段裁剪数据集,区域不小于480x480像素。 |
| 1200 | 500 | 606.96 MiB | IMG+TXT | 短边不超过1200像素的数据集。 |
| stage3-p480-1200 | 1257 | 1.23 GiB | IMG+TXT | 3阶段裁剪数据集,区域不小于480x480像素。 |
标签聚类结果
原始文本版本
| # | Samples | Img-1 | Img-2 | Img-3 | Img-4 | Img-5 | 标签 |
|---|---|---|---|---|---|---|---|
| 0 | 5 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, glasses, hair_ribbon, hat, looking_at_viewer, red-framed_eyewear, solo, blush, capelet, frills, skirt, white_thighhighs, book, open_mouth, simple_background, sitting, smile, under-rim_eyewear, adjusting_eyewear, holding, large_breasts, long_sleeves, shoes, white_background |
| 1 | 12 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, collarbone, large_breasts, looking_at_viewer, simple_background, solo, nipples, completely_nude, navel, white_background, cowboy_shot, open_mouth, closed_mouth, groin, stomach, sweat |
| 2 | 6 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, bare_shoulders, collarbone, looking_at_viewer, shirt, short_sleeves, solo, upper_body, white_background, blush, detached_sleeves, simple_background, closed_mouth, medium_breasts, open_mouth, see-through_sleeves |
| 3 | 9 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, bare_shoulders, blush, looking_at_viewer, short_sleeves, white_skirt, black_shirt, collarbone, solo, see-through, simple_background, white_background, detached_sleeves, floral_print, open_mouth, medium_breasts |
| 4 | 11 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, serafuku, solo, white_sailor_collar, hanasakigawa_school_uniform, looking_at_viewer, short_sleeves, blue_shirt, collarbone, pleated_skirt, simple_background, blue_neckerchief, white_skirt, closed_mouth, smile, white_background, white_shirt, open_mouth, upper_body |
| 5 | 25 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, hanasakigawa_school_uniform, solo, blush, white_sailor_collar, looking_at_viewer, red_ribbon, long_sleeves, upper_body, neck_ribbon, open_mouth, simple_background, white_background, sailor_dress, buttons, hair_between_eyes, smile, collarbone |
| 6 | 6 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, black_pantyhose, blush, cowboy_shot, hanasakigawa_school_uniform, long_sleeves, looking_at_viewer, neck_ribbon, red_ribbon, sailor_dress, solo, standing, brown_dress, collarbone, pleated_dress, white_sailor_collar, closed_mouth, double-breasted, simple_background, large_breasts, medium_breasts, smile, white_background |
| 7 | 11 | ![]() |
![]() |
![]() |
![]() |
![]() |
1boy, 1girl, hetero, nipples, solo_focus, blush, collarbone, large_breasts, open_mouth, penis, mosaic_censoring, navel, completely_nude, pussy, sex, sweat, cum, looking_at_viewer, paizuri, upper_body, vaginal |
| 8 | 5 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blue_sky, blush, cleavage, cloud, collarbone, day, looking_at_viewer, open_mouth, outdoors, solo, halterneck, large_breasts, ocean, bare_shoulders, beach, covered_nipples, long_sleeves, navel, white_bikini, blurry, medium_breasts, off_shoulder, open_jacket, purple_jacket, sand, standing, stomach, string_bikini, water |
| 9 | 10 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, solo, blue_shirt, long_sleeves, looking_at_viewer, necklace, collarbone, pom_pom_(clothes), brown_skirt, open_mouth, plaid_skirt, happy_birthday, simple_background, upper_body, vertical-striped_shirt, white_background |
| 10 | 8 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, looking_at_viewer, short_sleeves, solo, hair_bow, white_shirt, armband, collared_shirt, cosplay, pleated_skirt, simple_background, tokiwadai_school_uniform, brown_sweater_vest, open_mouth, ribbon, white_background, grey_skirt, medium_breasts, standing, :d, brown_hair, cowboy_shot, miniskirt, red_bow, safety_pin |
| 11 | 9 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, hair_ribbon, looking_at_viewer, smile, short_sleeves, solo, star_(symbol), upper_body, white_shirt, electric_guitar, holding_instrument, bracelet, collared_shirt, frills, playing_instrument, purple_bowtie, suspender_skirt |
| 12 | 5 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, earrings, frills, gloves, heart, looking_at_viewer, solo, brooch, hair_bow, red_bow, red_headwear, ribbon, blush, detached_collar, long_sleeves, open_mouth, upper_body, white_background, :d, ascot, cleavage, dress, hat_flower, jacket, light_brown_hair, medium_breasts, mini_top_hat, simple_background |
| 13 | 12 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, solo, floral_print, obi, long_sleeves, looking_at_viewer, hair_flower, open_mouth, wide_sleeves, purple_kimono, outdoors, :d, print_kimono, upper_body |
| 14 | 6 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, bell, blush, cleavage, frills, hat, solo, christmas, looking_at_viewer, red_dress, red_headwear, ribbon, star_(symbol), bare_shoulders, scarf, smile, black_thighhighs, medium_breasts, red_flower, simple_background, wrist_cuffs |
| 15 | 9 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, looking_at_viewer, solo, white_apron, frilled_apron, maid_headdress, open_mouth, maid_apron, black_dress, bow, enmaided, puffy_short_sleeves, simple_background, standing, white_background |
| 16 | 7 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, hair_flower, looking_at_viewer, solo, white_dress, holding_bouquet, white_gloves, bare_shoulders, bridal_veil, bride, collarbone, rose, strapless_dress, wedding_dress, :d, cleavage, medium_breasts, necklace, open_mouth, petals, purple_flower |
| 17 | 10 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, looking_at_viewer, solo, blush, red_capelet, animal_ears, braid, hooded_capelet, nail_polish, open_mouth, earrings, halloween_costume, hood_up, red_nails, choker, claw_pose, purple_bow, :d, frills, hair_bow, cleavage, little_red_riding_hood_(grimm)_(cosplay), striped, upper_body |
| 18 | 8 | ![]() |
![]() |
![]() |
![]() |
![]() |
1girl, blush, detached_collar, fake_animal_ears, rabbit_ears, solo, bare_shoulders, cleavage, looking_at_viewer, playboy_bunny, strapless_leotard, black_leotard, large_breasts, black_pantyhose, bowtie, hairband, medium_breasts, open_mouth, rabbit_tail, wrist_cuffs, simple_background, brown_pantyhose, cowboy_shot, fake_tail, ribbon, sitting, standing, white_background |
搜集汇总
数据集介绍

构建方式
本数据集聚焦于BanG Dream!角色市ヶ谷有咲,通过自动化爬取系统从Danbooru、Pixiv、Zerochan等多个图像平台收集原始图像,最终收录500张图片及其对应标签。数据集由DeepGHS团队开发,采用Waifuc框架进行数据采集与预处理。原始图像经过边缘对齐至1400像素的规范化处理,并进一步衍生出多种分辨率版本,包括短边不超过800像素和1200像素的压缩集,以及基于三阶段裁剪策略生成的480x480像素以上区域裁剪集,以满足不同训练需求。
特点
该数据集的核心特征在于其精细的标签体系与多版本结构。角色关键属性如金发、长发、刘海、发饰、双马尾、棕眼、侧发、胸部、黄眼等被明确标记并精简。数据集提供原始压缩包、多种分辨率图像及三阶段裁剪集,总计五个子集,其中裁剪集通过数据增强扩充至1257个样本。此外,数据集还包含标签聚类结果,将图像按照服装、场景等特征分为19个簇,便于研究者挖掘角色在不同装扮下的视觉模式。
使用方法
研究者可通过Hugging Face Hub直接下载数据集压缩包,或使用Waifuc库加载原始数据集。加载时需先通过huggingface_hub下载dataset-raw.zip,解压后利用LocalSource接口读取图像及其元数据,包括文件名和标签信息。数据集支持多分辨率选择,用户可根据模型输入尺寸灵活选用800、1200或三阶段裁剪版本。对于需要特定场景或服装组合的训练任务,可参考提供的聚类标签表快速筛选子集。
背景与挑战
背景概述
该数据集由CyberHarem团队于近期创建,聚焦于《BanG Dream!》中角色市ヶ谷有咲(ichigaya_arisa)的视觉表征。作为二次元文化中典型的金发双马尾少女形象,该角色在动漫与游戏领域拥有广泛受众。数据集共收录500张图像及其对应标签,图像来源涵盖Danbooru、Pixiv、Zerochan等知名插画平台,并通过DeepGHS团队开发的自动化爬取系统完成采集。核心标签如blonde_hair、twintails、yellow_eyes等被系统化裁剪,旨在为文本到图像生成任务提供高质量、高一致性的训练素材。该数据集的发布填补了特定动漫角色细粒度视觉数据集的空白,对推动二次元角色生成模型的发展具有示范意义。
当前挑战
当前数据集面临多维度挑战。在领域问题层面,文本到图像生成任务需应对角色外观的复杂多样性,如不同画风、服饰变换(校服、泳装、礼服等)及姿态差异,要求模型具备跨场景的泛化能力。构建过程中,数据来源的异构性带来标签一致性难题,不同平台标注规范各异,需通过自动化与人工校验结合的方式确保标签准确性。此外,图像分辨率与裁剪策略的权衡(如800px与1200px版本)影响模型训练效率与生成质量,而数据量仅500张的限制进一步加剧了过拟合风险。隐私与合规方面,部分内容涉及成人向标签(如nipples、completely_nude),需在开放使用与内容审核间寻求平衡。
常用场景
经典使用场景
该数据集专为文本到图像生成任务而设计,聚焦于《BanG Dream!》中的角色市ヶ谷有咲,包含500张经过精细标注的图像。其经典使用场景在于训练和微调基于扩散模型的图像生成器,如Stable Diffusion,以实现对特定动漫角色形象的高保真重建。通过提供多分辨率版本(如800px、1200px)及裁剪后的数据包,研究者能够灵活适配不同计算资源与模型架构,从而在角色一致性、风格迁移和细节保真度等维度上优化生成效果。
实际应用
在实际应用中,该数据集支撑了虚拟偶像创作、同人画作辅助生成以及游戏角色概念设计等场景。内容创作者可借助基于该数据集训练的模型,快速生成符合角色设定(如发型、服饰、配饰)的图像,大幅降低手工绘制成本。此外,其标签聚类结果(如校服、泳装、婚礼服等)为自动化图像编辑与风格化渲染提供了结构化先验,在数字娱乐与文创产业中具有显著的落地价值。
衍生相关工作
该数据集衍生了一系列经典工作,包括基于Waifuc框架的自动化图像采集与标注流程,以及多阶段裁剪(3-stage cropping)策略,后者显著提升了训练数据的质量与多样性。相关研究还探索了通过标签聚类挖掘角色外观模式的方法,为后续的细粒度属性编辑与跨模态检索任务奠定了基础。此外,该数据集已成为社区中动漫角色生成基准测试的重要组成,催生了针对长尾角色与复杂场景的生成模型改进方案。
以上内容由遇见数据集搜集并总结生成


































































































