Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Face Detection
Abstract Factory
Viola Jones
Integral Image
• An integral image is an algorithm for
quickly and efficiently generating the sum
of values in a rectangular subset of a grid
• The integral image can be computed from
an image using a few operations per pixel
Integral Image
Haar Features
• Once the integral image was computed,
any one of these Harr-like features can be
computed at any scale or location in
constant time.
Haar Features
Machine Learning
• AdaBoost => XML
• Cascade of Classifiers
AspectJ
• Profiler
– 6s/thread
• Logger
Face Recognition
Biometrics
Biometric characteristics:
● physiological
● behavioral
Why is it useful?
How the biometric system works?
● verification (one-to-one)
● identification (one-to-many)
Biometrics
Diagram of a biometric system
Stages of facial
recognition
1. Image capture
2. Face identification
3. Extraction of features to
generate a model
4. Comparing models
5. Declaration of identity
Algorithms
As researcher interest in face recognition continued, many
different algorithms were developed, three of which have
been well studied in face recognition literature:
● Principal Components Analysis (PCA)
● Linear Discriminant Analysis (LDA)
● Elastic Bunch Graph Matching (EBGM)
Method PCA Eigenfaces
Database
Database
Problems
• ● background
• ● lighting conditions (performance
• degrades with light changes)
• ● scale (performance decreases quickly
• with changes to the head size)
• ● orientation (perfomance decreases but
not as fast as with scale changes)
• ● less reliable than face detection (30-70%)
Faces
Distributed system
Time issues
Face detection – 6/7 seconds for an
average size photo
We want to be able to respond to multiple
face detection requests which come at the
same time
Don’t want our users to wait too much
Solution ?
Parallel/Distributed
Computing
JPPF
JPPF
OPEN SOURCE
easy programming model
fault-tolerance and self-repair capabilities
dynamically scalable on-demand
fine-grained monitoring and administration
JPPF architecture
Server – handles all tasks, knows all
active nodes, load balancing, queues
Node – executes tasks received from the
server
Client – issues tasks to the Server, waits
for responses
Demo
(hope this works)