MonoPerfCap: Human Performance Capture from Monocular Video

TL;DR

MonoPerfCap introduces a novel marker-less method for capturing 3D human performance from monocular video, effectively handling occlusions and non-rigid deformations for applications like video editing and free viewpoint viewing.

Contribution

It is the first approach to achieve temporally coherent 3D human performance capture from monocular video with general clothing, using a neural network-based pose detection and a batch-based reconstruction strategy.

Findings

Outperforms previous monocular methods in accuracy and robustness

Handles complex scenes with occlusions and non-rigid deformations

Enables applications like video editing and free viewpoint video

Abstract

We present the first marker-less approach for temporally coherent 3D performance capture of a human with general clothing from monocular video. Our approach reconstructs articulated human skeleton motion as well as medium-scale non-rigid surface deformations in general scenes. Human performance capture is a challenging problem due to the large range of articulation, potentially fast motion, and considerable non-rigid deformations, even from multi-view data. Reconstruction from monocular video alone is drastically more challenging, since strong occlusions and the inherent depth ambiguity lead to a highly ill-posed reconstruction problem. We tackle these challenges by a novel approach that employs sparse 2D and 3D human pose detections from a convolutional neural network using a batch-based pose estimation strategy. Joint recovery of per-batch motion allows to resolve the ambiguities of the monocular reconstruction problem based on a low dimensional trajectory subspace. In addition, we propose refinement of the surface geometry based on fully automatically extracted silhouettes to enable medium-scale non-rigid alignment. We demonstrate state-of-the-art performance capture results that enable exciting applications such as video editing and free viewpoint video, previously infeasible from monocular video. Our qualitative and quantitative evaluation demonstrates that our approach significantly outperforms previous monocular methods in terms of accuracy, robustness and scene complexity that can be handled.