|
|
Face Recognition Using Wavelet Decompositions From Different Pruning Algorithms
Fall
2005, Carnegie Mellon University
A comparison
is presented between two major pruning strategies
for face recognition using wavelet packets.
The first approach is top-down, beginning
with the image representation of the human
faces and decomposing the image selectively
based on correlation energy. The second approach
begins with the full wavelet decomposition,
designs a correlation filter for each subspace,
and prunes the tree based on classification
rate. Each technique has advantages: top-down
is better at verifying whether an image belongs
to a particular class while the full wavelet
decomposition is better at determining which
class an image belongs to out of a set of
different classes. We attempt to generate
trees that take advantage of both strengths
to see at what middle ground we maximize
effectiveness at image verification and image
classification, using faces as the test images.
Then, we apply the tree structure to sets
of human faces to determine whether the tree
structure is more broadly applicable.
|
|
 |
Vehicle Navigation Using 3D Visualization
Summer
2004, 2005, Robert Bosch Corporation - Research
Training Center (RTC)
Traditional
navigation visualization utilizes two
dimensional maps for road guidance or
arrow symbols for turn by turn information.
While the advantage of map views is supposed
to be the inherent understanding of the
surroundings, often these schematic line-drawing
bird’s eye views
are rather confusing than helpful because
they cannot provide an overview and an
appropriate level of detail in an area
of interest at the same time, i.e. the
user is forced to change between different
resolutions. In this project, the HMI
group makes use of 3D visualization to
display geo information to the customers.
I worked on the offline process of the
generation, classification, and retrieval
of the huge data set. I wrote some library
tools that have been incorporated in
the existing library on which the whole
system operates.
Proceed
to details
|
|
Illumination-Tolerance Face Verification
Spring
2005, Carnegie Mellon University
The
goal is to examine the recognition rate
of nine pattern recognition methods on
faces with varying illumination. The
same set of training images and test images
has been used for all methods that were
tested. The different methods analyzed
are in the following order: Global Principal
Component Analysis, Individual Principal
Component Analysis, Fisherface, 3D Linear
Subspace, Support Vector Machine, MACE,
UMACE, OTSDF and MACH correlation Filters.
In reality, we can not get a perfect
set of training images. During the recognition
process, light illuminates the face from
every angle and could affect the recognition
rate greatly. Therefore, we want to find
out which pattern recognition method
performs the best in the worst scenario.
We used the CMU PIE database with no background
light and chose a specific training set
for the entire analysis. Most face recognition
methods perform well with varying illumination
training set. We measure the performance
of each recognition algorithm by their
corresponding recognition rate. We also
have the ROC curve for each recognition
method. At low Probability of Imposter,
the higher the Probability of Authentic
is, the better the recognition method
is.
Proceed
to details
|
|
last
3 << >> next
3
|
|
