Faculty Candidate: Svetlana Lazebnik/Beckman Institute for Science & Tech. University of Illinois at Urbana-Champaign Local Semi-Local and Global Models for Texture Object and Scene Recognition in ACES 2.302

Contact Name: 
Jenna Whitney
Date: 
Mar 28, 2006 11:00am - 12:00pm


There is a signup schedule for this event.

Speaker Name/Affiliation: Svetlana Lazebnik/Beckman Institute for Science

& Tech. University of Illinois at Urbana-Champaign

Talk Title: Loca

l Semi-Local and Global Models for Texture Object and Scene Recognition
Date/Time: March 28 2006 at 11:00 a.m.

Coffee: 10:45 a.m.

Location: ACES 2.302

Host: Okan Arikan

Talk Abstract:<

br>I will present my work on recognizing materials scenes and
objects

in photographs --- key computer vision problems that
are made challengin

g by the seemingly limitless variability
of natural imagery. Currently

even the most advanced
recognition systems lack the geometric invariance
robustness
and flexibility to cope with the full range of this variab

ility.
To overcome these limitations I have developed several
approa

ches combining salient local image features with spatial
relations and d

iscriminative learning techniques. First
I will discuss a simple yet ef

fective orderless image
representation that was originally designed for

the problem of
recognizing images of textured surfaces subjected to view

point
changes and non-rigid deformations. In a large-scale comparativeevaluation this method has also performed well for object
categorizat

ion despite substantial clutter and occlusion. Next
I will discuss an e

xtension of this method that incorporates
global spatial information for
classification of natural scene
categories. Finally I will describe a

part-based object
recognition approach that supports the learning of rob

ust and
geometrically invariant object models from small sets of
unse

gmented cluttered training images. Baseline comparisons
show that each

of the proposed approaches is capable of
outperforming the state of the

art on challenging datasets.

Apart from my work on recognition I am
also interested in
acquiring high-fidelity 3D models of objects from ph

otographs
and video. In this area I have worked on image-based techniqu

es
for reconstructing 3D shapes from silhouettes and local texture
w

ith applications to 3D photography and video shot matching.