cs479/679 pattern recognition dr. george bebis
DESCRIPTION
Feature Selection Jain, A.K.; Duin , P.W.; Jianchang Mao, “Statistical pattern recognition: a review”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 22, no. 1, pp. 4-37, Jan. 2000. Section 4.2 Genetic Algorithms Chapter 7 ( Duda et al.) – Section 7.5. - PowerPoint PPT PresentationTRANSCRIPT
![Page 1: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/1.jpg)
Feature SelectionJain, A.K.; Duin, P.W.; Jianchang Mao, “Statistical pattern
recognition: a review”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 22, no. 1, pp. 4-37, Jan. 2000.
Section 4.2
Genetic AlgorithmsChapter 7 (Duda et al.) – Section 7.5
CS479/679 Pattern RecognitionDr. George Bebis
![Page 2: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/2.jpg)
Feature Selection• Given a set of n features, the role of feature
selection is to select a subset of d features (d < n) in order to minimize the classification error.
• Fundamentally different from dimensionality reduction (e.g., PCA or LDA) based on feature combinations (i.e., feature extraction).
dimensionality reduction
![Page 3: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/3.jpg)
Feature Selection vs Dimensionality Reduction
• Dimensionality Reduction– When classifying novel patterns, all features need to be computed.– The measurement units (length, weight, etc.) of the features are
lost.
• Feature Selection– When classifying novel patterns, only a small number of features
need to be computed (i.e., faster classification).– The measurement units (length, weight, etc.) of the features are
preserved.
![Page 4: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/4.jpg)
Feature Selection Steps
• Feature selection is an optimization problem.
– Step 1: Search the space of possible feature subsets.
– Step 2: Pick the subset that is optimal or near-optimal with respect to some objective function.
![Page 5: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/5.jpg)
Feature Selection Steps (cont’d)
Search strategies– Optimum– Heuristic– Randomized
Evaluation strategies- Filter methods - Wrapper methods
![Page 6: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/6.jpg)
Search Strategies• Assuming n features, an exhaustive search would
require:
– Examining all possible subsets of size d.
– Selecting the subset that performs the best according to the criterion function.
• The number of subsets grows combinatorially, making exhaustive search impractical.
• In practice, heuristics are used to speed-up search but they cannot guarantee optimality.
6
n
d
![Page 7: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/7.jpg)
Evaluation Strategies
• Filter Methods– Evaluation is independent
of the classification algorithm.
– The objective function evaluates feature subsets by their information content, typically interclass distance, statistical dependence or information-theoretic measures.
![Page 8: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/8.jpg)
Evaluation Strategies
• Wrapper Methods– Evaluation uses criteria
related to the classification algorithm.
– The objective function is a pattern classifier, which evaluates feature subsets by their predictive accuracy (recognition rate on test data) by statistical resampling or cross-validation.
![Page 9: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/9.jpg)
Filter vs Wrapper Approaches
![Page 10: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/10.jpg)
Filter vs Wrapper Approaches (cont’d)
![Page 11: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/11.jpg)
Naïve Search
• Sort the given n features in order of their probability of correct recognition.
• Select the top d features from this sorted list.
• Disadvantage– Feature correlation is not considered.– Best pair of features may not even contain the best individual
feature.
![Page 12: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/12.jpg)
Sequential forward selection (SFS)(heuristic search)
• First, the best single feature is selected (i.e., using some criterion function).
• Then, pairs of features are formed using one of the remaining features and this best feature, and the best pair is selected.
• Next, triplets of features are formed using one of the remaining features and these two best features, and the best triplet is selected.
• This procedure continues until a predefined number of features are selected.
12
SFS performsbest when the optimal subset issmall.
![Page 13: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/13.jpg)
Example
13
Results of sequential forward feature selection for classification of a satellite image using 28 features. x-axis shows the classification accuracy (%) and y-axis shows the features added at each iteration (the first iteration is at the bottom). The highest accuracy value is shown with a star.
features added at each iteration
![Page 14: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/14.jpg)
Sequential backward selection (SBS) (heuristic search)
• First, the criterion function is computed for all n features.
• Then, each feature is deleted one at a time, the criterion function is computed for all subsets with n-1 features, and the worst feature is discarded.
• Next, each feature among the remaining n-1 is deleted one at a time, and the worst feature is discarded to form a subset with n-2 features.
• This procedure continues until a predefined number of features are left.
14
SBS performsbest when the optimal subset islarge.
![Page 15: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/15.jpg)
Example
15
Results of sequential backward feature selection for classification of a satellite image using 28 features. x-axis shows the classification accuracy (%) and y-axis shows the features removed at each iteration (the first iteration is at the top). The highest accuracy value is shown with a star.
features removed at each iteration
![Page 16: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/16.jpg)
Bidirectional Search (BDS)
• BDS applies SFS and SBS simultaneously:– SFS is performed from the empty set– SBS is performed from the full set
• To guarantee that SFS and SBS converge to the same solution:– Features already selected by SFS are not
removed by SBS– Features already removed by SBS are not
added by SFS
![Page 17: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/17.jpg)
SFS and SBS
• The main disadvantage of SFS is that it is unable to remove features that become obsolete after the addition of other features.
• The main limitation of SBS is its inability to reevaluate the usefulness of a feature after it has been discarded.
![Page 18: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/18.jpg)
“Plus-L, minus-R” selection (LRS)
• A generalization of SFS and SBS– If L>R, LRS starts from the empty set and:
• Repeatedly add L features • Repeatedly remove R features
– If L<R, LRS starts from the full set and:• Repeatedly removes R features• Repeatedly add L features
Its main limitation is the lack of a theory to help predict the optimal values of L and R.
![Page 19: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/19.jpg)
Sequential floating selection (SFFS and SFBS)
• An extension to LRS with flexible backtracking capabilities– Rather than fixing the values of L and R, floating methods
determine these values from the data.– The dimensionality of the subset during the search can be
thought to be “floating” up and down
• There are two floating methods:– Sequential floating forward selection (SFFS) – Sequential floating backward selection (SFBS)
P. Pudil, J. Novovicova, J. Kittler, Floating search methods in feature selection, Pattern Recognition Lett. 15 (1994) 1119–1125.
![Page 20: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/20.jpg)
Sequential floating forward selection (SFFS)
• Sequential floating forward selection (SFFS) starts from the empty set.
• After each forward step, SFFS performs backward steps as long as the objective function increases.
![Page 21: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/21.jpg)
Sequential floating backward selection (SFBS)
• Sequential floating backward selection (SFBS) starts from the full set.
• After each backward step, SFBS performs forward steps as long as the objective function increases.
• SFBS algorithm is analogous to SFFS.
![Page 22: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/22.jpg)
Feature Selection using Genetic Algorithms (GAs)
(randomized search)
ClassifierFeature Subset
Pre-Processing
Feature Extraction
Feature Selection
(GA)
Feature Subset
GAs provide a simple, general, and powerful framework
for feature selection.
![Page 23: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/23.jpg)
Genetic Algorithms (GAs)
• What are GAs?– An optimization technique for searching very large spaces.– Inspired by the biological mechanisms of natural selection
and reproduction.
• Main characteristics of GAs– Global optimization technique.– Searches probabilistically using a population of structures,
each properly encoded as a string of symbols.– Uses an objective function, not derivatives.
![Page 24: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/24.jpg)
• Encoding scheme– Represent solutions in parameter space as finite length
strings (chromosomes) over some finite set of symbols.
Encoding
(11,6,9) (1011_ 0110 _1001) (101101101001)
![Page 25: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/25.jpg)
•Fitness function– Evaluates the goodness of a solution.
Fitness Evaluation
( ( ))Fitness f decode chromosome
![Page 26: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/26.jpg)
Searching
10010110… 10010110… 01100010… 01100010… 10100100... 10100100… 10010010… 01111001… 01111101… 10011101…
SelectionCrossover
Mutation
Current Generation Next Generation
![Page 27: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/27.jpg)
Selection• Probabilistically filters out solutions that perform poorly,
choosing high performance solutions to exploit.
0.1
0.9
0.01
0.01
1001 1001
1101 1101
1000 1101
0001 1101
fitness
![Page 28: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/28.jpg)
• Explore new solutions.• Crossover: information exchange between strings.
• Mutation: restore lost genetic material.
Crossover and Mutation
10011110 10011010
mutated bit
10011110 10010010
10110010 10111110
![Page 29: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/29.jpg)
Feature Selection Using GAs
• Binary encoding
• Fitness evaluation (to be maximized)
Feature#1 Feature#N
fitness=104accuracy + 0.4 #zeros
Classification accuracy using avalidation set
number offeatures
![Page 30: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/30.jpg)
Case Study 1: Gender Classification
• Determine the gender of a subject from facial images.– Challenges: race, age, facial expression, hair style, etc.
Z. Sun, G. Bebis, X. Yuan, and S. Louis, "Genetic Feature Subset Selection for Gender Classification: A Comparison Study", IEEE Workshop on Applications of Computer Vision, pp. 165-170, Orlando, December 2002.
![Page 31: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/31.jpg)
Feature Extraction Using PCA
• PCA maps the data in a lower-dimensional space using a linear transformation.
• The columns of the projection matrix are the “best” eigenvectors (i.e., eigenfaces) of the covariance matrix of the data.
![Page 32: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/32.jpg)
Which eigenvectors encode mostly gender information?
EV#1 EV#2 EV#3 EV#4 EV#5 EV#6
EV#8 EV#10 EV#12 EV#14 EV#19 EV#20
IDEA: Use GAs to search the space of eigenvectors!
![Page 33: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/33.jpg)
Dataset• 400 frontal images from 400 different people
– 200 male, 200 female
– Different races, lighting conditions, and facial expressions
• Images were registered and normalized– No hair information
– Normalized to account for different lighting conditions
![Page 34: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/34.jpg)
Experiments• Classifiers
– LDA
– Bayes classifier
– Neural Networks (NNs)
– Support Vector Machines (SVMs)
• Three-fold cross validation– Training set: 75% of the data
– Validation set: 12.5% of the data
– Test set: 12.5% of the data
• Comparison with SFBS
![Page 35: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/35.jpg)
Error Rates
ERM: error rate using top eigenvectors
ERG: error rate using GA selected eigenvectors
17.7%
11.3%
22.4%
13.3% 14.2%
9% 8.9%
4.7%
6.7%
![Page 36: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/36.jpg)
Ratio of Features - Information Kept
RN: percentage of eigenvectors in the feature subset selected.
RI: percentage of information contained in the eigenvector subset selected.
17.6%
38%
13.3%
31%36.4%
61.2%
8.4%
32.4%
42.8%
69.0%
![Page 37: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/37.jpg)
Histograms of Selected Eigenvectors
(a) LDA (b) Bayes
(d) SVMs(c) NN
![Page 38: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/38.jpg)
Reconstructed Images
Reconstructed faces using GA-selected EVs do not contain information about identity but disclose strong gender information!
Original images
Using top 30 EVs
Using EVs selected by SVM+GA
Using EVs selected by NN+GA
![Page 39: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/39.jpg)
Comparison with SFBS
Original images
Top 30 EVs
EVs selected by SVM+GA
EVs selectedby SVM+SFBS
![Page 40: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/40.jpg)
Case Study 2: Vehicle Detectionlow light camera
rear views
Non-vehicles class muchlarger than vehicle class.
Z. Sun, G. Bebis, and R. Miller, "Object Detection Using Feature Subset Selection", Pattern Recognition, vol. 37, pp. 2165-2176, 2004.
Ford Motor Company
![Page 41: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/41.jpg)
Which eigenvectors encode themost important vehicle features?
![Page 42: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/42.jpg)
Experiments• Training data set (collected in Fall 2001)
2102 images (1051 vehicles and 1051 non-vehicles)
• Test data sets (collected in Summer 2001)
231 images (vehicles and non-vehicles)
• SVM for classification• Three-fold cross-validation• Comparison with SFBS
![Page 43: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/43.jpg)
Error Rate
6.49%
![Page 44: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/44.jpg)
Histograms of Selected Eigenvectors
SFBS-SVM GA-SVM
Number of eigenvectors selected by SBFS: 87 (43.5% information)
Number of eigenvectors selected by GA: 46 (23% information)
![Page 45: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/45.jpg)
Vehicle Detection
Original
Top 50 EVs
EVs selectedby SFBS
Evs selectedby GAs
Reconstructed images using the selected feature subsets. - Lighting differences have been disregarded by the GA approach.
![Page 46: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/46.jpg)
Case Study 3:Fusion of Visible-Thermal IR Imagery
for Face Recognition
G. Bebis, A. Gyaourova, S. Singh, and I. Pavlidis, "Face Recognition by Fusing Thermal Infrared and Visible Imagery", Image and Vision Computing, vol. 24, no. 7, pp. 727-742, 2006.
![Page 47: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/47.jpg)
Visible vs Thermal IR• Visible spectrum
– High resolution, less sensitive to the presence of eyeglasses.
– Sensitive to changes in illumination and facial expression.
• Thermal IR spectrum– Robust to illumination changes and facial expressions.– Low resolution, sensitive to air currents, face heat patterns,
aging, and the presence of eyeglasses (i.e., glass is opaque to thermal IR).
![Page 48: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/48.jpg)
How should we fuse information from visible and thermal IR?
Feature Extraction
Fusion UsingGenetic Algorithms
Reconstruct Image
FusedImage
We have tested two feature extraction methods: - PCA and Wavelets
![Page 49: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/49.jpg)
Dataset
Co-registered thermal infrared and visible images.
Variations among images:
- Eyeglasses
- Illumination (front, left, right)
- Facial expression (smile, frown, surprise and vowels)
Equinox database
![Page 50: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/50.jpg)
Eyeglasses/Illumination Tests
ELG
ELnG
EFnG
EFG
Frontal Illuminatio
n
Glasses
No Glasses
EnG
EG
Lateral illuminatio
n
DATA
![Page 51: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/51.jpg)
Experiments
EG ELG EFG EnG ELnG EFnG
EG
ELG
EFG
EnG
ELnG
EFnG
no glasses
glasses
glasses
no glasse
s
database
test
Eyeglasses Tests
Illumination Tests
![Page 52: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/52.jpg)
Results
0102030405060708090
100
ELG
-ELn
G
EFG
-EFn
G
ELnG
-ELG
EFnG
-EFG
ELG
-EFG
EFG
-ELG
ELnG
-EFn
G
EFnG
-ELn
G
ELG
-EFn
G
EFG
-ELn
G
ELnG
-EFG
EFnG
-ELG
EG-E
nG
EnG
-EG
ELnG
-EG
EFnG
-EG
EG-E
LnG
EG-E
FnG
ELG
-EnG
EFG
-EnG
EnG
-ELG
EnG
-EFG
Presence ofeyeglasses
Illumination direction Eyeglasses andillumination direction
Eyeglasses and mixed illumnation
infrared
vis ible
fused
![Page 53: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/53.jpg)
Facial Expression Tests
EA
Smile, Frown & Surprise
Frontal Illuminatio
n
Lateral illuminatio
n
EF
EL
VA
Speaking Vowels
Frontal Illuminatio
n
VF
VL
![Page 54: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/54.jpg)
Results
0102030405060708090
100
EF-V
F
VF-
EF
EL-V
L
VL-
EL
EF-E
L
EL-E
F
VF-
VL
VL-
VF
EF-V
L
VL-
EF
VF-
EL
EL-V
F
EA-V
A
VA
-EA
Facial expression (emotionsvs. pronunciation of vocals)
Illumination direction Facial expression andillumination direction
Facialexpressionand mixedillumination
infrared
visible
fused
![Page 55: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/55.jpg)
Overall Results
![Page 56: CS479/679 Pattern Recognition Dr. George Bebis](https://reader036.vdocuments.mx/reader036/viewer/2022081417/56815a11550346895dc759a5/html5/thumbnails/56.jpg)
Overall Results