Monocular Camera Dense 3D Reconstruction of a Dynamically Deforming Surface

Speaker:        Prof Hongdong LI
                Australian National University

Title:          "Monocular Camera Dense 3D Reconstruction of a Dynamically
                Deforming Surface"

Date:           Thursday, 6 June 2019

Time:           11:00am - 12 noon

Venue:          Room 2463 (via lift no. 25/26), HKUST

Abstract:

In this talk, I will describe some of our recent work dense 3D geometry
reconstruction for a dynamically deformable shape or scene using a single
monocular camera.  We aim to show that it is possible to recover the 3D
geometry of a dynamic deformable surface under mild assumption or
condition of the scene or shape. Traditional methods for dynamic 3D
reconstruction often require stereo camera setup, or assume the scene
(with deformable object) follows some low-rank linear model. Our method
removes such restrictive requirement. We test our method on standard
benchmarks datasets, including "KITTI" for autonomous driving and
"Sintel"--an open source animated movie.  If times allows, I will also
cover a work on 3D dynamic human pose recovery using repetitive structure.


***************
Biography:

Hongdong Li is currently a Reader with the Computer Vision Group of ANU
(Australian National University). He is a Chief Investigator for the
Australia Centre for Robotic Vision. His research interests include 3D
Computer Vision, Camera Calibration, Robot navigation as well as
applications of global optimization in vision. He worked on the "Australia
Bionic Eyes" project in 2008-2010 with NICTA.  He served as the Area Chair
in recent year CVPR, ICCV, ECCV, BMVC and 3DV.  He is an Associate Editor
on the Editorial Board for IEEE Transactions on PAMI (T-PAMI), and Program
Co-Chair for the Asian Conference on Computer Vision 2018 and Co-General
Chair for ACCV 2022 in Macao China.  Jointly with his PhD students he has
won several most prestigious paper awards in computer vision, including
the CVPR Best Paper Award, IEEE ICIP Best Student Paper Award, IEEE-ICIP
Best Student Paper Award, as well as the Marr Prize-Honourable Mention at
ICCV 2017.