Documente Academic
Documente Profesional
Documente Cultură
Real-time Face
Recognition
By: Ali Zakeri
Tafresh University
Winter 2014
Outlines
History
Applications
Steps and methods
Problems and approaches
New concept consequence scope
History
1st semi-automated system
Problem: Manually
computation of
measurements and
locations
1990s
1970s
1988
1991
Applications
Security
Fight terrorism
Find fugitives
ATM
Sporting events
Home access (no keys or passwords)
Any other application that would want personal identification
Face
Detection
Face
Normalization
Face
Identification
Approaches
A
PCA
B
LDA
EBGM
Principal
Component
Analysis
Linear
Discrimination
Analysis
LBP
Elastic
Bunch
Graph
Matching
Local
Binary
Pattern
Combination
with Other
Approaches
Aspects
Enhancement
of Its
Robustness
Selection of Its
Neighborhood
Extension to
3-D Data
Discriminative Capability
H. Yang et al. proposed an improved LBP (ILBP)
Discriminative Capability
Jin et al. introduced an extended LBP(ELBP)
Guo et al. proposed a complete LBP (CLBP)
10
Robustness
LBP is sensitive to noise
Tan and Triggs proposed local ternary patterns
(LTPs)
11
Neighborhood
Neighboring pixels in the original LBP are defined on a
circle
Liao and Chung :main reason to obtain rotation invariance
for texture description
This is not suitable for all applications
They proposed elongated LBP
12
3-D LBP
Two difficulties:
I. Equidistant sampling on a sphere is a difficult job
II. It is also difficult to set an order to those sampling point
. Volume LBP (VLBP or 3-D-LBP) was introduced by Zhao
and Pietikainen
13
3-D LBP
14
Combined LBP
A set of approaches was proposed to combine
Gabor wavelets
Gabor-wavelet- and LBP-based features are
mutually complementary
Some approaches:
Multiresolution histograms of local variation
patterns (MHLVPs)
local Gabor binary pattern histogram (LGBPH)
15
Problems
Accurate and real-time face
recognition due to:
1.
2.
3.
4.
5.
6.
Lighting variation
Orientation variation (face angle)
Size variation
Large database
Processor intensive
Time requirements
16
Accurate Face
Recognition
Ahonen et al. introduced LBP in face recognition
Characteristics:
Nearest neighbor (NN) classifier
Their approach outperforms the PCA, EBGM and
the Bayesian classifier
On all four probe sets of the FERET database
17
Accurate Face
Recognition
Zhang et al. [24] introduced MHLVP for face
recognition
They achieved 95.9% accuracy on the fc set with
illumination changes
They employed LGBPH which is similar to MHLVP
Very promising result on FERET and AR dataset
18
Accurate Face
Recognition
Hadid et al. introduced VLBP
To extract local facial dynamics for
spatiotemporal face recognition from video
sequences
AdaBoost was applied to learn the specific facial
dynamics
Ignoring intrapersonal temporal information
Superior performances on various databases:
MoBo (97.9%), Honda/UCSD (96.0%), and CRIM
(98.5%)
19
Real-time Face
Recognition
Methods
1
CPU
Parallel
implementation
on GPU, FPGA,
GPU languages
CUDA
CTM
Brook
OPENCL
20
Real-time Face
Recognition
Altaf Ahmad Huqqani
et al. :
2 novel parallel
training approaches
For Backpropagation
neural networks for
face recognition
OpenMP and CUDA
21
Real-time Face
Recognition
CYN Dwith and Rathna.G.N fulfilled:
LBP based face recognition algorithm on GPU
They applied parallelizing by OPENBCL
22
Real-time Face
Recognition
Relative improvement in
computational time
Computational efficiency
of GPU vs. CPU
23
NEW CONCEPT
Y. Guo et al. developed a feature extraction
method for texture description
Simultaneously
1. considering the robustness
2. Discriminative power
3. representation capability of features
24
NEW CONCEPT
25
SCOPE
Phase
Phase 11
new concept
+ LBP
variations
Phase
Phase 22
Phase
Phase 33
Implementati Parallel
on on face
processing on
recognition
GPU
te
me
l-ti
rea
ura
acc
26
References
[1] CYN Dwith, Dr. Rathna.G.N, Parallel Implementation Of LBP Based Face Recognition
On Gpu Using Opencl,2012 13th International Conference on Parallel and Distributed
Computing, Applications and Technologies, IEEE_06589372.
[2] CYN Dwith, Dr. Rathna.G.N, Parallel Computing for Accelerated Texture Classification
with Local Binary Pattern Descriptors using OpenCL, International Journal of Computer
Applications (0975 8887) Volume 64 No.1, February 2013.
[3] Yimo Guo, GuoyingZhao, MattiPietikainen, Discriminative features for texture
description, SciVerse ScienceDirect on Pattern Recognition 45(2012)38343843.
[4] Di Huang, Student Member, IEEE, Caifeng Shan, Member, IEEE, Mohsen Ardabilian,
Yunhong Wang, Member, IEEE, and Liming Chen, Member, IEEE, Local Binary Patterns and
Its Application to Facial Image Analysis: A Survey, IEEE TRANSACTIONS ON SYSTEMS,
MAN, AND CYBERNETICSPART C: APPLICATIONS AND REVIEWS, VOL. 41, NO. 6,
NOVEMBER 2011.
[5]Altaf Ahmad Huqqania , Erich Schikutaa, Sicen Yea, Peng Chena, Multicore and GPU
Parallelization of Neural Networks for Face Recognition, International Conference on
Computational Science, ICCS 2013.
27
Thank You !
Thanks of your attention.
Is there any question?
Company
LOGO