Poses of People in Art: A Dataset for Human Pose Estimation in Digital Art History

Schneider, Stefanie und Vollmer, Ricarda (2024): Poses of People in Art: A Dataset for Human Pose Estimation in Digital Art History. In: ACM Journal on Computing and Cultural Heritage, Vol. 17, No. 4, 61

Abstract

With the Poses of People in Art dataset, we introduce the first openly licensed dataset for estimating human poses in art and validating human pose estimators. It consists of 2,454 images from 22 art-historical depiction styles, including those that have increasingly moved away from lifelike representations of the body since the 19th century. A total of 10,749 human f igures are enclosed by rectangular bounding boxes, with a maximum of 4 per image labeled by up to 17 keypoints. For machine learning purposes, the dataset is divided into three subsets—training, validation, and test—that follow the JSON-based Microsoft Common Objects in Context (COCO) format, respectively. Each image annotation provides metadata from the online visual art encyclopedia WikiArt, in addition to mandatory fields. In this article, we report on the acquisition and constitution of the dataset, address various application scenarios, and discuss the prospects for a digitally supported art history. We show that the dataset allows for the study of body phenomena in art, whether on the level of individual figures, which can thus be captured in their subtleties, or entire figure constellations, whose position or distance to each other is considered.

• Item Type: Journal article
• Faculties: History and Art History > Department of Art History > Art History
• Subjects: 700 Arts and recreation > 700 Arts
• ISSN: 1556-4673
• Language: English
• DFG: Gefördert durch die Deutsche Forschungsgemeinschaft (DFG) - 415796915