face recognition demo
TRANSCRIPT
-
8/18/2019 Face Recognition Demo
1/16
FACE RECOGNITION
PROJECT
-
8/18/2019 Face Recognition Demo
2/16
Outline
1. Face image cropping and preprocessing
2. Appearance-based face recognition methods•
Principal Component Analysis (PCA)• Linear Discriminant Analysis (LDA)
• Locality Preserving Projections (LPP)
• K-nearest neighbor as the classification method
3. Experimental Results
4. Demonstration
-
8/18/2019 Face Recognition Demo
3/16
Face image cropping and preprocessing
• Face Detection with OpenCV Library
• Normalization
• Cropping
• Grayscale conversion
• Fixing of locations of eyes (by OpenCV)
• Fixing to the same resolution
• Preprocessing
• Histogram Equalization
-
8/18/2019 Face Recognition Demo
4/16
Our Dataset
-
8/18/2019 Face Recognition Demo
5/16
Appearance-based methods
• PCA
• dimensionality reduction method
• produces a compact representation
• LDA
• supervised learning algorithm
• find a subspace which separate different classes of objects
• LPP• considers the manifold structure for face analysis
-
8/18/2019 Face Recognition Demo
6/16
1. Training
2. Recognizing
Reference: [1]
-
8/18/2019 Face Recognition Demo
7/16
Eigenfaces
Fisherfaces
Laplacianfaces
Eigenfaces, Fisherfaces & Laplacianfaces
PCA
LDA
LPP
-
8/18/2019 Face Recognition Demo
8/16
PCA Reconstruction
Original Face d = 2 d = 6 d = 10 d = 20 d = 40 d =60
Original Face d = 2 d = 6 d = 10 d = 20 d = 40 d =60
-
8/18/2019 Face Recognition Demo
9/16
Experimental Results
1. PCA with different energy percentage
2. Different metric types
3. Different partitions for validation
4. Different K for K-nearest neighbor
5. LPP with different PCA processing
-
8/18/2019 Face Recognition Demo
10/16
PCA with different energy percentage
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.5 0.6 0.7 0.8 0.9 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99
own dataset
ORL dataset
A c c u r a c y
Energy Percentage
-
8/18/2019 Face Recognition Demo
11/16
Different metric types
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Cosine Dissimilarity L1 Distance Euclidean Distance
PCA
LDA
LPP
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Cosinedissimilarity
L1 distance Euclideandistance
PCA
LDA
LPP
ORL Dataset Our Dataset
Accuracy Accuracy
-
8/18/2019 Face Recognition Demo
12/16
Different partitions for validation
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
3 4 5
PCA
LDA
LPP
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
3 4 5
PCA
LDA
LPP
Training NumberTraining Number
Accuracy Accuracy
ORL Dataset Our Dataset
-
8/18/2019 Face Recognition Demo
13/16
Different K for K-nearest neighbor
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1 2 3
PCA
LDA
LPP
KNN_K0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1 2 3
PCA
LDA
LPP
KNN_K
Accuracy Accuracy
ORL Dataset Our Dataset
-
8/18/2019 Face Recognition Demo
14/16
LPP with different PCA processing
0.7
0.75
0.8
0.85
0.9
0.95
40 60 80 100
LPP
PCA
Number of
Components
Accuracy
-
8/18/2019 Face Recognition Demo
15/16
Conclusion
• PCA performs the best on ORL dataset
• LDA performs the best on our dataset
•
Out dataset includes lots of variations
• LPP does not perform as expected
• This method still could achieve good results with proper setting
• A more complex method calculating weight matrix may help
-
8/18/2019 Face Recognition Demo
16/16
References
[1] Delac, K., Grgic, M., & Grgic, S. (2005). Independent comparative study ofPCA, ICA, and LDA on the FERET data set. International Journal of ImagingSystems and Technology , 15 (5), 252-260.
[2] Turk, M., & Pentland, A. (1991). Eigenfaces for recognition. Journal ofcognitive neuroscience, 3(1), 71-86.
[3] Belhumeur, P. N., Hespanha, J. P., & Kriegman, D. J. (1997). Eigenfaces vs.fisherfaces: Recognition using class specific linear projection. Pattern Analysisand Machine Intelligence, IEEE Transactions on, 19(7), 711-720.
[4] He, X., Yan, S., Hu, Y., Niyogi, P., & Zhang, H. J. (2005). Face recognitionusing laplacianfaces. Pattern Analysis and Machine Intelligence, IEEETransactions on, 27 (3), 328-340.
[5] Viola, P., & Jones, M. J. (2004). Robust real-time face detection.International journal of computer vision, 57 (2), 137-154.