jaddai/openart-portraits-classical
收藏资源简介:
OpenArt — Portraits & the Classical Figure(openart-portraits-classical)是OpenArt开放公共领域艺术数据集系列中的肖像古典主题集合。该数据集包含28,011件作品(13,868幅绘画/插图、13,970张拍摄对象照片、173件未分类),每件作品都配有结构化的视觉语言模型(VLM)字幕,以及媒介、归属和铭文元数据。数据集涵盖全范围媒介中的人物形象和肖像艺术,包括绘画和素描肖像、摄影肖像、肖像纺织品和服装、雕刻和雕塑 likeness。2D肖像与拍摄/对象肖像的比例接近均等。数据来源于大都会艺术博物馆、荷兰国家博物馆、克利夫兰艺术博物馆、芝加哥艺术博物馆、史密森尼学会、Europeana、维基媒体等主要博物馆和图书馆的开放访问API。与仅包含绘画的OpenBrush语料库不同,每个OpenArt主题集合都是混合媒介的:既包含2D艺术(绘画、版画、素描),也包含三维对象的照片(雕塑、陶瓷、金属制品、纺织品)。数据集通过“brand”列进行分割,以便用户根据需要选择特定部分。关键特征包括:所有图像均通过“rights_status: safe”检查,为公共领域;提供结构化v2字幕(每个图像9个语义部分,与OpenBrush模式相同);品牌分割为“openbrush”(2D艺术)和“openartifacts”(拍摄对象);媒介标记和归属感知,其中6,648行包含命名艺术家,5,600行经过签名验证;采用CC0-1.0许可,为公共领域奉献,无保留权利;每行提供完整来源信息,包括“source”、“landing_page”和“sha256”。数据集结构包括图像、宽度和高度、品牌、主题、标签、字幕部分(如主题、动作、设置、情绪、风格描述、灯光、颜色、构图)、完整字幕、媒介、艺术家、签名、签名验证、归属置信度、铭文文本、铭文可读性和验证、标题、权利信息以及来源字段。字幕生成采用双模型流程:首先使用Gemma 4 31B生成完整字幕、身份保留主题描述、铭文阅读和路由;然后使用Gemini 3 Flash进行验证,重新确认媒介、归属和铭文信息。数据集适用于图像生成训练、VLM微调、分类和跨媒介研究。
OpenArt — Portraits & the Classical Figure (openart-portraits-classical) is the portraits classical subject collection of the OpenArt family of open, public-domain art datasets. It contains 28,011 works (13,868 paintings/illustrations, 13,970 photographed objects, 173 unclassified), each paired with a structured VLM caption plus medium, attribution and inscription metadata. The dataset covers the human figure and portraiture across the full range of media, including painted and drawn portraits alongside photographic portraits, portrait textiles and costume, engraved and sculpted likenesses, with a near-even split between 2-D portraiture and photographed/object portraiture. It draws from the open-access APIs of major museums and libraries such as the Metropolitan Museum of Art, Rijksmuseum, Cleveland Museum of Art, Art Institute of Chicago, Smithsonian, Europeana, Wikimedia and others. Unlike the painting-only OpenBrush corpus, each OpenArt subject collection is mixed-medium, containing both 2-D art (paintings, prints, drawings) and photographs of three-dimensional objects (sculpture, ceramics, metalwork, textiles). A brand column splits every row for user selection. Key features include: 28,011 public-domain images passing a rights_status: safe gate; structured v2 captions with 9 semantic sections per image (same schema as OpenBrush); brand split into openbrush (13,868, 2-D art) and openartifacts (13,970, photographed objects); medium-tagged and attribution-aware with 6,648 rows carrying a named artist and 5,600 signature-verified; CC0-1.0 public-domain dedication; full provenance per row with source, landing_page, and sha256. The dataset structure includes fields for image, width, height, brand, theme, tags, caption sections (subject, action, setting, mood, style_description, lighting, color, composition), caption_full, medium, artist, signatures, signature_verified, attribution_confidence, inscription_text, inscription_legible, inscription_verified, title, rights, rights_status, license_class, landing_page, source, and sha256. Captions are generated via a two-model pipeline: Gemma 4 31B for captioning and routing, and Gemini 3 Flash for verification of medium, attribution, and inscription. The dataset is suitable for image-generation training, VLM fine-tuning, classification, and cross-medium study.




