BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Model-Based Hand Tracking with Texture\, Shading and Self-occlusio
 ns - Martin De La Gorce\, Ecole des Ponts Paris Tech
DTSTART:20151210T100000Z
DTEND:20151210T110000Z
UID:TALK62895@talks.cam.ac.uk
CONTACT:44515
DESCRIPTION:I will first present a model-based approach to 3D hand trackin
 g from monocular video. The 3D hand pose\, the hand texture and the illumi
 nant are dynamically estimated through minimisation of an objective functi
 on. Derived from an inverse problem formulation\, the objective function e
 nables explicit use of texture temporal continuity and shading information
 \, while handling important self-occlusions and time-varying illumination.
  The minimisation is done efficiently using a quasi-Newton method\, for wh
 ich we propose a rigorous derivation of the objective function gradient. P
 articular attention is given to terms related to the change of visibility 
 near self-occlusion boundaries that are neglected in existing formulations
 . In doing so we introduce new occlusion forces and show that using all gr
 adient terms greatly improves the performance of the method. Experimental 
 results demonstrate the potential of the formulation.\n\nI will then brief
 ly present a more recent work for the 3D reconstruction of a piecewise-pla
 nar watertight surface from range images\, typically indoor laser scans wi
 th millions of points. We formulate the surface reconstruction problem as 
 a integer programming problem where we assign a binary label to elements o
 f a partition of the scene volume that is based on detected and hypothesis
 ed planes. Our major contributions is a regularisation of the reconstructe
 d surface w.r.t. the length of edges and the number of corners. Compared t
 o classical area-based regularisation\, it better captures surface complex
 ity and is therefore better suited for man-made environments\, such as bui
 ldings. We approximately solve the integer program using a simple continuo
 us relaxation. Experiments show that it is fast and reaches near-optimal s
 olution.\n\nReferences:\nM. de la Gorce\, N. Paragios and David Fleet. Mod
 el-Based Hand Tracking with Texture\, Shading and Self-occlusions. IEEE Co
 nference in Computer Vision and Pattern Recognition (CVPR)\, Anchorage 200
 8\nhttp://www.gigovery.com/research_page/publications/CVPR_2008_de_La_Gorc
 e.pdf  \n\nAlexandre Boulch\, Martin de La Gorce\, Renaud Marlet Piecewise
 -Planar 3D Reconstruction with Edge and Corner Regularization. Comput. Gra
 ph. Forum \nhttp://onlinelibrary.wiley.com/doi/10.1111/cgf.12431/pdf  \n\n
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR