General Automatic Human Shape and Motion Capture Using Volumetric Contour Cues

TL;DR

This paper introduces a fully automatic multi-view method for human shape and motion capture that jointly estimates personalized skeleton, shape, appearance, and motion from outdoor videos without manual initialization.

Contribution

It presents a novel automatic approach that combines a new image formation model, a Gaussian density shape approximation, and a statistical human body model for joint pose and shape estimation.

Findings

Works with outdoor multi-view videos with few cameras

Automatically creates rigged models including skeleton and shape

Achieves accurate motion capture without manual intervention

Abstract

Markerless motion capture algorithms require a 3D body with properly personalized skeleton dimension and/or body shape and appearance to successfully track a person. Unfortunately, many tracking methods consider model personalization a different problem and use manual or semi-automatic model initialization, which greatly reduces applicability. In this paper, we propose a fully automatic algorithm that jointly creates a rigged actor model commonly used for animation - skeleton, volumetric shape, appearance, and optionally a body surface - and estimates the actor's motion from multi-view video input only. The approach is rigorously designed to work on footage of general outdoor scenes recorded with very few cameras and without background subtraction. Our method uses a new image formation model with analytic visibility and analytically differentiable alignment energy. For reconstruction, 3D body shape is approximated as Gaussian density field. For pose and shape estimation, we minimize a new edge-based alignment energy inspired by volume raycasting in an absorbing medium. We further propose a new statistical human body model that represents the body surface, volumetric Gaussian density, as well as variability in skeleton shape. Given any multi-view sequence, our method jointly optimizes the pose and shape parameters of this model fully automatically in a spatiotemporal way.