postPerspective

Capturing Humans

Wednesday, 27 July, 3:45 pm - 5:55 pm, Anaheim Convention Center, Ballroom D
Session Chair: Richard Szeliski, Facebook, University of Washington

Fusion4D: Real-Time Performance Capture of Challenging Scenes

This demonstration of the first real-time multi-view performance capture system, with quality on par with offline methods that are several orders of magnitude slower achieves 4D spatio-temporal reconstruction of challenging scenes with robust topology changes and fast motions.

Mingsong Dou
Microsoft Research

Sameh Khamis
Microsoft Research

Yury Degtyarev
Microsoft Research

Philip Davidson
Microsoft Research

Sean Ryan Fanello
Microsoft Research

Adarsh Kowdle
Microsoft Research

Sergio Orts Escolano
Microsoft Research

Christoph Rhemann
Microsoft Research

David Kim
Microsoft Research

Jonathan Taylor
Microsoft Research

Pushmeet Kohli
Microsoft Research

Vladimir Tankovich
Microsoft Research

Shahram Izadi
Microsoft Research, Microsoft Corporation

An Anatomically Constrained Local Deformation Model for Monocular Face Capture

This new anatomically constrained local 3D face model can be fit to 2D motion data such as optical flow from monocular video, marker-based motion capture, or artist sketches. In contrast to global blendshape models, the local deformation model provides more expressiveness at a reduced model size.

Chenglei Wu
ETH Zürich

Derek Bradley
Disney Research

Markus Gross
Disney Research, ETH Zürich

Thabo Beeler
The Walt Disney Company, Disney Research

AutoHair: Fully Automatic Hair Modeling From A Single Image

Introducing the first fully automatic method for 3D hair modeling from a single image. The method's efficacy and robustness is demonstrated on internet photos, resulting in a database of 50K 3D hair models and a corresponding hairstyle space that covers a wide range of real-world hairstyles.

Menglei Chai
Zhejiang University

Tianjia Shao
Zhejiang University

Hongzhi Wu
Zhejiang University

Yanlin Weng
Zhejiang University

Kun Zhou
Zhejiang University

Lightweight Eye Capture Using a Parametric Model

A data-driven parametric model of the human eye, built from high-quality 3D reconstructions, and an image-based fitting method to reconstruct eyes from sparse data, such as single images or artistic drawings, making high-quality eye reconstruction simple and widely available.

Pascal Berard
ETH Zürich, Disney Research

Derek Bradley
Disney Research

Markus Gross
ETH Zürich, Disney Research

Thabo Beeler
Disney Research

Reconstruction of Personalized 3D Face Rigs From Monocular Video

An automatic approach to creating fully parametrized, high-quality 3D face rigs from monocular video, including legacy footage, and how they can be used in relevant facial animation and video editing tasks.

Pablo Garrido
Max-Planck-Institut für Informatik

Michael Zollhöfer
Max-Planck-Institut für Informatik

Dan Casas
Max-Planck-Institut für Informatik

Levi Valgaerts
Max-Planck-Institut für Informatik

Kiran Varanasi
Technicolor SA

Patrick Perez
Technicolor SA

Christian Theobalt
Max-Planck-Institut für Informatik

Real-Time 3D Eye Gaze Animation Using a Single RGB Camera

This paper presents the first real-time facial performance capture system that simultaneously captures the coordinated movements of 3D head poses, facial expression deformation, and eye gaze using a single video camera.

Congyi Wang
Chinese Academy of Sciences

Fuhao Shi
Texas A&M University

Shihong Xia
Chinese Academy of Sciences

Jinxiang Chai
Texas A & M University