imet-open-access
收藏资源简介:
iMet开放访问数据集(大都会艺术博物馆藏品多标签属性)是一个用于多标签图像分类任务的数据集。该数据集包含来自纽约大都会艺术博物馆的藏品图像,每张图像带有多个属性标签,这些标签涵盖了国家(如阿富汗、中国、法国等)、文化(如古希腊、埃及、伊斯兰等)以及媒介(如金、银、油画、陶瓷等)等类别,共计数百个细粒度标签。数据规模在10万到100万之间,分为训练集、验证集和测试集。该数据集适用于多标签分类、细粒度分类、长尾分布学习以及艺术/博物馆领域的图像理解研究。
The iMet Open Access Dataset (Metropolitan Museum of Art Collection Multi-label Attributes) is a dataset for multi-label image classification. It contains images of artworks from the Metropolitan Museum of Art in New York, each annotated with multiple attribute labels covering categories such as countries (e.g., Afghanistan, China, France), cultures (e.g., Ancient Greek, Egyptian, Islamic), and media (e.g., gold, silver, oil painting, ceramic), totaling hundreds of fine-grained labels. The dataset size ranges from 100k to 1M images, split into training, validation, and test sets. It is suitable for research on multi-label classification, fine-grained classification, long-tail distribution learning, and image understanding in the art/museum domain.




