Gamasutra

If you missed SIGGRAPH 2017 watch a selection of recorded Live Streaming Sessions.

You are here

Technical Papers

Learning & Analysis for Geometry

Hewlett Packard

ACM DL

Tuesday, 1 August, 2:00 pm - 3:30 pm, Los Angeles Convention Center, Los Angeles Convention Center - Room 152
Session Chair: Justin Solomon, Massachusetts Institute of Technology

Learning Hierarchical Shape Segmentation and Labeling From Online Repositories

This method for converting geometric shapes into hierarchically segmented parts with part labels trains category-specific models from the scene graphs and part names that accompany 3D shapes in public repositories, which enables detection of hierarchies in novel 3D shapes and their constituent parts.

Li Yi
Stanford University

Leonidas Guibas
Stanford University

Aaron Hertzmann
Adobe Research

Vladimir Kim
Adobe Research

Hao Su
Stanford University

Ersin Yumer
Adobe Research

Convolutional Neural Networks on Surfaces via Seamless Toric Covers

This paper presents a method for applying deep learning to shapes using a global, seamless parameterization of a planar flat-torus, for which the convolution operator is well defined. As a result, the standard deep-learning framework can be readily applied for learning semantic, high-level properties of the shape.

Haggai Maron
Weizmann Institute of Science

Meirav Galun
Weizmann Institute of Science

Noam Aigerman
Weizmann Institute of Science

Miri Trope
Weizmann Institute of Science

Nadav Dym
Weizmann Institute of Science

Ersin Yumer
Adobe Research

Vladimir Kim
Adobe Research

Yaron Lipman
Weizmann Institute of Science

O-CNN: Octree-B ased Convolutional Neural Networks for 3D Shape Analysis

O-CNN, an octree based convolutional neural network (CNN) for 3D shape analysis, is memory- and computation-efficient, and it makes 3D CNN for high-resolution 3D models possible.

Peng-Shuai Wang
Microsoft Research Asia, Tsinghua University

Yang Liu
Microsoft Research Asia

Yu-Xiao Guo
University of Electronic Science and Technology of China, Microsoft Research Asia

Chun-Yu Sun
Microsoft Research Asia, Tsinghua University

Xin Tong
Microsoft Research Asia

ClothCap: Seamless 4D Clothing Capture and Retargetting

ClothCap takes 3D sequences of clothed people in motion, segments the clothing from the body into pieces, aligns it in time, relates the pieces to the naked body underneath, and enables retargeting to new body shapes. The realism exceeds previous methods, providing a path towards virtual try-on without simulation.

Gerard Pons-Moll
Max Planck Institute for Intelligent Systems, Max-Planck-Institut für Informatik

Sergi Pujades
Max Planck Institute for Intelligent Systems, Max-Planck-Institut für Informatik

Sonny Hu
Body Labs

Michael Black
Max Planck Institute for Intelligent Systems, Max-Planck-Institut für Informatik