Estimation of Human Body Shape in Motion with Wide Clothing
Estimating 3D human body shape in motion from a sequence of unstructured oriented 3D point clouds is important for many applications. We propose the first automatic method to solve this problem that works in the presence of loose clothing. The problem is
- PDF / 1,730,737 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 25 Downloads / 226 Views
2
Inria Grenoble Rhˆ one-Alpes, Grenoble, France {jinlong.yang,jean-sebastien.franco, franck.hetroy,stefanie.wuhrer}@inria.fr Laboratoire Jean Kuntzmann, Universit´e Grenoble Alpes, Grenoble, France
Abstract. Estimating 3D human body shape in motion from a sequence of unstructured oriented 3D point clouds is important for many applications. We propose the first automatic method to solve this problem that works in the presence of loose clothing. The problem is formulated as an optimization problem that solves for identity and posture parameters in a shape space capturing likely body shape variations. The automation is achieved by leveraging a recent robust pose detection method [1]. To account for clothing, we take advantage of motion cues by encouraging the estimated body shape to be inside the observations. The method is evaluated on a new benchmark containing different subjects, motions, and clothing styles that allows to quantitatively measure the accuracy of body shape estimates. Furthermore, we compare our results to existing methods that require manual input and demonstrate that results of similar visual quality can be obtained.
Keywords: Human body modeling Statistical shape space
1
·
Shape and motion estimation
·
Introduction
Estimating 3D human body shape in motion is important for applications ranging from virtual change rooms to security. While it is currently possible to effectively track the surface of the clothing of dressed humans in motion [2] or to accurately track body shape and posture of humans dressed in tight clothing [3], it remains impossible to automatically estimate the 3D body shape in motion for humans captured in loose clothing. Given an input motion sequence of raw 3D meshes or oriented point clouds (with unknown correspondence information) showing a dressed person, the goal Electronic supplementary material The online version of this chapter (doi:10. 1007/978-3-319-46493-0 27) contains supplementary material, which is available to authorized users. c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part IV, LNCS 9908, pp. 439–454, 2016. DOI: 10.1007/978-3-319-46493-0 27
440
J. Yang et al.
of this work is to estimate the body shape and motion of this person. Existing techniques to solve this problem are either not designed to work in the presence of loose clothing [4,5] or require manual initialization for the pose [6,7], which limits their use in general scenarios. The reason is that wide clothing leads to strong variations of the acquired surface that is challenging to handle automatically. We propose an automatic framework that allows to estimate the human body shape and motion that is robust to the presence of loose clothing. Existing methods that estimate human body shape based on an input motion sequence of 3D meshes or oriented point clouds use a shape space that models human body shape variations caused by different identities and postures as prior. Such a prior allows to reduce the search space to likely body shapes and postures. Prior work
Data Loading...
