Semi-supervised Human Pose Estimation in Art-historical Images

TL;DR

This paper introduces a semi-supervised method for human pose estimation in art-historical images, addressing domain differences with a new dataset and outperforming existing approaches that rely on pre-trained models or style transfer.

Contribution

It presents a novel semi-supervised learning approach and a new domain-specific dataset for pose estimation in art-historical images, improving over prior methods.

Findings

Significantly better pose estimation results than pre-trained or style transfer methods.

Introduction of a new annotated art-historical dataset.

Demonstrated effectiveness of semi-supervised learning in a challenging domain.

Abstract

Gesture as language of non-verbal communication has been theoretically established since the 17th century. However, its relevance for the visual arts has been expressed only sporadically. This may be primarily due to the sheer overwhelming amount of data that traditionally had to be processed by hand. With the steady progress of digitization, though, a growing number of historical artifacts have been indexed and made available to the public, creating a need for automatic retrieval of art-historical motifs with similar body constellations or poses. Since the domain of art differs significantly from existing real-world data sets for human pose estimation due to its style variance, this presents new challenges. In this paper, we propose a novel approach to estimate human poses in art-historical images. In contrast to previous work that attempts to bridge the domain gap with pre-trained models or through style transfer, we suggest semi-supervised learning for both object and keypoint detection. Furthermore, we introduce a novel domain-specific art data set that includes both bounding box and keypoint annotations of human figures. Our approach achieves significantly better results than methods that use pre-trained models or style transfer.