Conference 3 Dec - 6 Dec Exhibition 4 Dec - 6 Dec


    Technical Papers

    01 Full Conference1 - Full Conference One Day



    Capturing Everything

    Saturday, 06 December

    14:15 - 16:00

    Sweet Osmanthus Hall

    Automatic Acquisition of High-fidelity Facial Performances Using Monocular Videos

    We present a video-based motion capture technique that automatically and accurately captures high-fidelity facial performances using uncontrolled monocular videos (e.g., Internet videos). We show our system advances the state of the art in video-based facial performance capture by comparing against alternative methods.

    Fuhao Shi, Texas A&M University
    Muscle Wu, Microsoft Research Asia
    Xin Tong, Microsoft Research Asia
    Jinxiang Chai, Texas A&M University

    High-Quality Capture of Eyes

    We present the first method to capture human eyes in high-quality, including the sclera, cornea and iris, showing fine scale details that make each individual eye unique. We also reconstruct iris deformation during pupil dilation and demonstrate applications of iris animations for visual effects.

    Pascal BĂ©rard, ETH Zurich and Disney Research Zurich
    Derek Bradley, Disney Research Zurich
    Thabo Beeler, Disney Research Zurich
    Maurizio Nitti, Disney Research Zurich
    Markus Gross, ETH Zurich and Disney Research Zurich

    Dynamic Hair Capture Using Spacetime Optimization

    We present a dynamic hair capture system for reconstructing realistic hair motions from multiple synchronized video sequences. This is achieved by using a novel hair motion tracking algorithm and formulating the global hair reconstruction as a spacetime optimization problem.

    Zexiang Xu, Beihang University
    Hsiang-Tao Wu, Microsoft Research Asia
    Lvdi Wang, Microsoft Research Asia
    Changxi Zheng, Columbia University
    Xin Tong, Microsoft Research Asia
    Yue Qi, Beijing University of Aeronautics and Astronautics

    Capturing Braided Hairstyles

    Braided hairstyles exhibit complex intertwining structures which are difficult to reconstruct using state-of-the-art hair capture techniques. We present a data-driven framework to faithfully reconstruct braided hairstyles by leveraging structure information provided by a procedurally created database.

    Liwen Hu, University of Southern California
    Chongyang Ma, University of Southern California
    Linjie Luo, Adobe Research
    Li-Yi Wei, The University of Hong Kong
    Hao Li, University of Southern California

    Back to Top
    Go Back