V4ldeLund/nationalmuseet-open-images

该数据集是一个从Nationalmuseet Samlinger Online独立收集的研究数据集，包含元数据和可选的WebDataset图像分片，涵盖许可类型为Public Domain、CC-BY和No known rights的Nationalmuseet资产记录。数据集主要用于丹麦语CLIP训练，其中preferred_title、preferred_description和preferred_text字段优先使用丹麦语（da-DK），英语作为备选。数据集包含151,287个可访问图像样本，分为公共领域、无已知权利和CC-BY许可，并提供了多种文件格式如Parquet、JSON和WebDataset shards，以支持图像分类、图像到文本和零样本图像分类任务。

This dataset is an independently harvested research dataset from Nationalmuseet Samlinger Online. It contains metadata and optionally WebDataset image shards for Nationalmuseet asset records whose rights.license is one of: Public Domain, CC-BY, or No known rights. The dataset is designed for Danish CLIP training, with Danish-first fields (da-DK) used for preferred_title, preferred_description, and preferred_text when available, and English as a fallback. It includes 151,287 accessible image samples across different license buckets, and provides files in formats such as Parquet, JSON, and WebDataset shards for tasks like image-classification, image-to-text, and zero-shot-image-classification.