Shape and Machine Learning

    Thursday, 21 November

    14:15 - 16:00

    Convention Hall B

    Fine-Grained Semi-Supervised Labeling of Large Shape Collections

    We introduce a multi-label semi-supervised approach that takes as input a large shape collection of a given category with associated sparse and noisy labels, and outputs cleaned and complete labels for each shape. Experimental results show that this method outperforms state-of-the-art semi-supervised learning techniques.

    Qi-xing Huang, Stanford University
    Hao Su, Stanford University
    Leonidas Guibas, Stanford University

    Efficient Penetration Depth Approximation using Active Learning

    Efficient PD approximation algorithm

    Reliable PD computation based on translation and/or rotational motion

    Interactive performance on non-convex and non-manifold rigid models

    Jia Pan, University of North Carolina (UNC) Chapel Hill
    Xinyu Zhang,University of North Carolina (UNC) Chapel Hill
    Dinesh Manocha, University of North Carolina (UNC) Chapel Hill

    Projective Analysis for 3D Shape Segmentation

    We introduce projective analysis for semantic segmentation and labelling of 3D shapes. The analysis treats an input 3D shape as a collection of 2D projections, labels each projection by transferring knowledge from existing labelled images, and back-projects and fuses the labellings on the 3D shape.

    Yunhai Wang, Shenzhen Institute of Advanced Technology
    Minglun Gong, Memorial University of Newfoundland
    Tianhua Wang, Jilin University
    Daniel Cohen-Or, Tel Aviv University
    Hao Zhang, Simon Fraser University
    Baoquan Chen, Shenzhen Institute of Advanced Technology

    3D Wikipedia: Using online text to automatically label and navigate reconstructed geometry

    Given a reference text, such as Wikipedia, and the site name, we automatically create a labeled 3D reconstruction, with objects in the model linked to where they appear in the text. Moreover, we have built a user interface that enables coordinated browsing of the text with the visualization

    Bryan C. Russell, Intel Labs
    Ricardo Martin-Brualla, University Of Washington
    Daniel Butler, University Of Washington
    Steve Seitz, University Of Washington
    Luke Zettlemoyer, University Of Washington
    Ricardo Martin-Brualla, University of Washington