3D Representation Learning for Visual Computing

(CSE Colloquium Lecture Series)

Speaker: Hao Su, Ph.D. candidate, Computer Science, Stanford University
Date: Monday, April 24, 2017
Time: 11:00am
Location: Room 1202, CSE Building

Title: 3D Representation Learning for Visual Computing

Abstract:  Among all digital representations we have for real physical objects, 3D is arguably the most expressive encoding. 3D representations allow storage and manipulation of high-level information (e.g. semantics, affordances, function) as well as low-level features (e.g. appearance, materials) about the object. How much we can understand and transform the 3D world is thus largely determined by the performance of algorithms that analyze and create 3D data. While 3D visual computing has predominantly focused on single 3D models or small model collections, the amount of accessible 3D models has increased by several orders of magnitude during the past few years. This significant growth pushes us to redefine 3D visual computing from the perspective of big 3D data.

In this talk, I will discuss a series of works on data-driven 3D visual computing. These include:  constructing an information-rich large-scale 3D model repository (ShapeNet), generating synthetic data for supervising neural networks (RenderForCNN), and learning end-to-end neural networks for analysis and synthesis of 3D geometries (PointNet and Point Set Generation Network). Under the guiding principle of using large-scale 3D data for representation learning, my efforts have led to top-performing algorithms for pure-3D data processing, as well as 3D-assisted semantic, geometric and physical property inference from 2D images. I will conclude my talk by describing several promising directions for future research.

Bio: Hao Su is currently a Ph.D. candidate in the Computer Science Department of Stanford University. He is a member of Stanford AI Lab and Geometric Computing Lab. Hao’s research interests are broad, spanning computer vision, computer graphics, machine learning, and robotics. He is particularly interested in deep learning for 3D data understanding and interconnecting 3D data with other modalities such as images and texts. Hao has published papers at CVPR, ICCV, NIPS, ICML, SIGGRAPH, SIGGRAPH Asia, VLDB, SIGSPATIAL, IJCV, etc. He is currently a student lead of the ShapeNet team. He has served as the chair of multiple international conferences and workshops (Co-chair of CVPR’15 workshop, Co-chair of ICCV’15 workshop, Co-chair of ECCV’16 workshop, Publication Chair of 3DVision'16, Program Chair of 3DVision'17). He is also an invited speaker at NIPS’16 workshop and 3DV'16 workshop on 3D deep learning.

Related Research Publications:

Faculty Host: Manmohan Chandraker (mkchandraker@eng.ucsd.edu)