+ All Categories
Home > Documents > EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62...

EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62...

Date post: 13-Dec-2015
Category:
Upload: dina-pearson
View: 238 times
Download: 1 times
Share this document with a friend
Popular Tags:
40
EE4-62 MLCV Lecture 13-14 Face Recognition Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV
Transcript
Page 1: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

1

EE4-62 MLCV

Lecture 13-14Face Recognition – Subspace/Manifold Learning

Tae-Kyun Kim

EE4-62 MLCV

Page 2: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

2

EE4-62 MLCVFace Image Tagging and Retrieval

• Face tagging at commercial weblogs

• Key issues– User interaction for face tags– Representation of a long- time

accumulated data– Online and efficient learning

• Active research area in Face Recognition Test and MPEG-7 for face image retrieval and automatic passport control• Our proposal promoted to

MPEG7 ISO/IEC standard

Page 3: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

3

EE4-62 MLCV

Principal Component Analysis (PCA)- Maximum Variance Formulation of PCA- Minimum-error formulation of PCA

- Probabilistic PCA

Page 4: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

4

Maximum Variance Formulation of PCA

Page 5: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

5

Page 6: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

6

Page 7: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

7

Page 8: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

8

Page 9: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

9

Minimum-error formulation of PCA

Page 10: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

10

Page 11: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

11

Page 12: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

12

Page 13: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

13

Page 14: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

14

Page 15: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

15

EE4-62 MLCV

Applications of PCA to Face Recognition

Page 16: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

16

EE4-62 MLCV

(Recap) Geometrical interpretation of PCA• Principal components are the vectors in the direction of the

maximum variance of the projection samples.

• Each two-dimensional data point is transformed to a single variable z1 representing the projection of the data point onto the eigenvector u1.

• The data points projected onto u1 has the max variance.• Infer the inherent structure of high dimensional data.• The intrinsic dimensionality of data is much smaller.

• For given 2D data points, u1 and u2 are found as PCs

Page 17: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

17

Eigenfaces• Collect a set of face images• Normalize for scale, orientation (using eye locations)

• Construct the covariance matrix and obtain eigenvectors

w

h

D=wh

NDRX

,...,1

1 xxXXXN

S T

MDRUUSU ,

Page 18: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

18

Eigenfaces• Project data onto the

subspace

• Reconstruction is obtained as

• Use the distance to the subspace for face recognition

DMRZXUZ NMT ,,

UZXUzuzxM

iii

~,~

1

x~||~|| xx

EE4-62 MLCV

x

Page 19: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

19

Page 20: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

20

EE4-62 MLCV

Matlab Demos – Face Recognition by PCA

Page 21: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

21

• Face Images• Eigen-vectors and Eigen-value plot• Face image reconstruction• Projection coefficients (visualisation of high-

dimensional data)• Face recognition

Page 22: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

22

Probabilistic PCA

• A subspace is spanned by the orthonormal basis (eigenvectors computed from covariance matrix)

• Can interpret each observation with a generative model

• Estimate (approximately) the probability of generating each observation with Gaussian distribution,

PCA: uniform prior on the subspace PPCA: Gaussian dist.

EE4-62 MLCV

Page 23: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

23

Continuous Latent Variables

Page 24: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

24

Page 25: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

25

Page 26: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

26

Probabilistic PCAEE4-62 MLCV

Page 27: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

27

Page 28: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

28

Page 29: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

29

Page 30: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

30

Page 31: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

31

Maximum likelihood PCA

Page 32: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

32

Page 33: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

33

Page 34: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

34

Page 35: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

35

Page 36: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

36

EE4-62 MLCV

Limitations of PCA

Page 37: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

37

Unsupervised learning

PCA vs LDA (Linear Discriminant Analysis)

Page 38: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

38

Linear model

Linear Manifold = Subspace Nonlinear

Manifold

EE4-62 MLCV

PCA vs Kernel PCA

Page 39: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

39

Gaussian Distribution Assumption

IC1

IC2

PC1

PC2

PCA vs ICA (Independent Component Analysis)

Page 40: EE4-62 MLCV Lecture 13-14 Face Recognition – Subspace/Manifold Learning Tae-Kyun Kim 1 EE4-62 MLCV.

EE4-62 MLCV

40

EE4-62 MLCV

(also by ICA)


Recommended