Multi-view Pictorial Structures for 3D Human Pose Estimation

Abstract

Pictorial structure models are the de facto standard for 2D human pose estimation. Numerous refinements and improvements have been proposed such as discriminatively trained body part detectors, flexible body models, and local and global mixtures. While these techniques allow to achieve state-of-the-art performance for 2D pose estimation, they have not yet been extended to enable pose estimation in 3D. This paper thus proposes a multi-view pictorial structures model that builds on recent advances in 2D pose estimation and incorporates evidence across multiple viewpoints to allow for robust 3D pose estimation. We evaluate our multi-view pictorial structures approach on the HumanEva-I and MPII Cooking dataset. In comparison to related work for 3D pose estimation our approach achieves similar or better results while operating on single-frames only and not relying on activity specific motion models or tracking. Notably, our approach outperforms state-of-the-art for activities with more complex motions.

Multiple Human Pose Estimation with Temporally Consistent 3D Pictorial Structures

“…We use the joint detector from [2] and learn the body prior using the ground-truth data from the KTH Multiview Football II dataset [9]. The evaluation is divided into three tasks: analysis of the state space, evaluation on the human detection and the body pose estimation.…”

“…The edges of the graph denote relation between the body joints. In comparison to the original pictorial structure model, we have interchanged the notion of the body parts with the joints to avoid the foreshortening effect [2]. Let the random variable Y i denote a joint location in the global 3D coordinate system, i.e.,…”

“…In order to be computationally efficient, we discretise the 3D space using 2D body joint detectors. We first sample 2D joint locations for each view using a body joint detector [2]. Next, we create the 3D locations by triangulation of the corresponding 2D body joints detected in all combinations of view pairs.…”

“…Pictorial structures is the most common part-based model for estimating the 2D body pose of single human [3,11,13]. The model has been extended to the 3D space, in order to cope with mutli-view camera setups as well [2,9,16]. Recently, pictorial structures have been successfully modelled for multiple human 3D pose estimation [6].…”

Using The Pictorial Structures in 3D Human Body Pose Estimation

“…So he applied the 2D pictorial structure models and estimated the 2D pose then the 3D pose is obtained by triangulation. (Amin et al, 2013).…”