| Date: Wednesday, December
10, 2003 Time:
11:00 AM
Location: Partners I - Room # 1101
Speaker: Dr. Washington
Mio
Department of Mathematics
Florida State University,
Tallahassee, FL
Title:
Shape Analysis for Image Understanding
Abstract:
We discuss a new framework for
the representation and quantitative analysis of planar shapes and
applications to the automated recognition and classification of objects in
digital images. Shape extraction is particularly challenging in imagery
involving partial occlusions of objects, noise, or low contrast. We take a
Bayesian approach to the problem, which requires the development of
image models encoding prior knowledge of shapes. This involves the problem
of clustering shapes and "learning" probability models for observed shapes
in addition to considerations of pixel models.
Date: Friday, December 12, 2003
Time:
12:00 PM
Location: EGRC building - Room # 136
Speaker: Dr.
Michael Kirby
Department
of Mathematics
Colorado State University, Fort
Collins, CO
Title:
Geometrical Structure in Large Data Sets
Abstract:
Some new tools for
characterizing information in large data sets will be presented. The
approach taken is highly geometrical with an emphasis on identifying and
representing coherent structure in the data. The subspace approach is
contrasted with the submanifold approach and applications are given that
illustrate the importance of each. A variety of data sets will be considered
from faces under varying illumination to numerical simulations of
hurricanes. The construction of low-dimensional dynamical models will also
be addressed.
|
