BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A Framework for Elastic Shape Analysis of Objects - Anuj Srivastav
 a\, Florida State University
DTSTART:20150417T150000Z
DTEND:20150417T160000Z
UID:TALK58853@talks.cam.ac.uk
CONTACT:20082
DESCRIPTION:Statistical analysis of shapes of objects is an important topi
 c area. Here tools are developed on a foundation of appropriate mathematic
 al representations and shape metrics.\nWhile a large body of prior work is
  based on registered landmarks\, more complex data demands more general so
 lutions.\nIn this talk we will look at some recent developments\, with a s
 pecial focus on the registration problem.\nRegistration is the process of 
 identifying points across objects.  It not only preserves important\nstruc
 tures in the data but also leads to more parsimonious statistical models f
 or capturing shape\nvariability. In case of parametrized curves and surfac
 es\, the registration step is akin to finding\noptimal parametrizations. T
 aking three fundamentally different examples: (1) real-valued function dat
 a\,\n(2) parametrized curves in Euclidean spaces\, and (3) parametrized su
 rfaces in R3\, I will describe\na Riemannian framework that achieves the f
 ollowing goals. It provides an analysis of shapes of objects\nthat is inva
 riant to standard similarity transformations and\, additionally\, to param
 etrizations\nof these objects. This framework\, called elastic shape analy
 sis\, incorporates an optimal registration\nof points across objects while
  simultaneously providing proper metrics\, geodesics\, and sample statisti
 cs\nof shapes. These sample statistics are further useful in statistical m
 odeling of shapes in different shape classes. I will demonstrate these ide
 as using applications from medical image analysis\, protein structure anal
 ysis\, 3D face recognition\, and human activity recognition in videos.
LOCATION:MR12\,  Centre for Mathematical Sciences\, Wilberforce Road\, Cam
 bridge
END:VEVENT
END:VCALENDAR