SniffyArt

SniffyArt数据集由弗里德里希-亚历山大大学埃尔兰根模式识别实验室创建，包含441幅历史艺术作品中的1941个人物，每个个体配有紧密的边界框、17个姿态关键点和手势标签。数据集通过合并五个独立的关键点注释集来确保高质量的人体姿态估计关键点。该数据集旨在通过结合姿态关键点和人物框，推动历史艺术作品中人体手势和嗅觉维度的分析，为多任务方法的研究奠定基础。

SniffyArt dataset was developed by the Pattern Recognition Lab at Friedrich-Alexander University Erlangen-Nuremberg. It contains 1941 human figures from 441 historical artworks, with each figure paired with tight bounding boxes, 17 pose keypoints and gesture labels. To ensure high-quality human pose estimation keypoints, the dataset integrates five independent keypoint annotation sets. This dataset aims to advance the analysis of human gestures and olfactory dimensions in historical artworks by combining pose keypoints and human bounding boxes, laying a solid foundation for research on multi-task methods.