KaoKore

KaoKore数据集是由ROIS-DS人文开放数据中心创建的，专注于从古代日本艺术作品中提取面部表情。该数据集包含5552张从16世纪至17世纪日本艺术品中提取的面部图像，每张图像均附有性别和社会地位等多标签信息。数据集的创建过程涉及利用国际图像互操作性框架（IIIF）进行大规模数字化和图像共享，确保了数据的可互操作性。KaoKore数据集不仅支持图像分类研究，还为艺术创作和生成模型提供了丰富的素材，旨在促进机器学习与日本艺术历史研究的交叉融合，解决艺术风格比较和艺术品作者鉴定等问题。

The KaoKore dataset was created by the ROIS-DS Center for Open Data in the Humanities, focusing on extracting facial expressions from ancient Japanese artworks. This dataset contains 5,552 facial images extracted from Japanese artworks dating from the 16th to 17th centuries, with each image accompanied by multi-label information such as gender and social status. The dataset creation process involved leveraging the International Image Interoperability Framework (IIIF) for large-scale digitization and image sharing, ensuring data interoperability. The KaoKore dataset not only supports image classification research but also provides rich resources for artistic creation and generative models, aiming to promote interdisciplinary integration between machine learning and Japanese art history studies, and address issues such as artistic style comparison and artwork author attribution.