face recognition using pca and neural network ppt................................................................................................................................................................................................................................
3. face recognition
What is face recognition?
face recognition system
face information
for each person
in a fixed domain
a person whose face is
in the input image
4. related researches
Template matching
very simple, very sensitive to noise
PCA(associative memory)
use PCA to extract important features and to reduce the dimension
of the input data
Image compression(back propagation network)
use compression network to extract important features and to
reduce the dimension of the input data
this is my approach
Others
extract visual feature like edges, shape of each components, etc.
5. My system
Object
face recognition for the face image domain in which there are some
small variations
face Image domain
15 persons
12 images per person
variation such as expression, direction of light, noise, glasses
cen gla hap lef nog noi
nor rig sad sle sur win
7. My system
Normalization
original image : 320*243
normalized image : 32*24
block size : 10*10
averaging the intensity of pixels in block
compression network
input = raw data of image
output = raw data of image ( same as input )
32*24 - 40 - 32*24 structure
backpropagation learning algorithm
8. My system
recognition network
input = values of hidden layer of compression network
output = a person whose face is in the image
40 - 10 - 15 structure
backpropagation learning algorithm
9. Experiment 1
Compression network
training data
nor images of 15 persons
2000 trainings, learning rate = 0.005
result(sample)
cen gla hap lef nog
noi
nor
rig sad sle sur win
Training
image
10. Experiment 2
Recognition network
use the result of experiment 1
training data
nor images of 15 persons
2000 trainings, learning rate = 0.05
result
total face image 180 ( including the training data)
the rate of correct recognition = 133/180*100 = 73.9%
the distribution of errors
nor cen gla hap lef nog noi rig sad sle sur win
0 6 3 3 12 2 0 12 2 2 2 3
11. Experiment 3
Recognition network
not use the result of experiment 1
assign random 40 dimensional key to each image
training data
nor image of 15 persons
2000 trainings, learning rate = 0.05
result
total face image 180 ( including the training data)
the rate of correct recognition =12/180*100 = 6.7%
the distribution of errors
nor cen gla hap lef nog noi rig sad sle sur win
14 14 14 14 14 14 14 14 14 14 14 14
12. Experiment 4
Recognition network
not use compression network
use raw image(32*24) for training and test
training data
nor image of 15 persons
2000 trainings, learning rate = 0.05
32*24 - 30-15 structure
result
total 180 images, the rate of correct recognition = 6.1%
the distribution of errors
nor cen gla hap lef nog noi rig sad sle sur win
14 14 14 15 14 14 13 15 14 14 14 14
13. Analysis
The hidden of compression network
encode the inputs in a smaller dimensional subspace that retains
most of the important information
if the hidden units are linear, the best solution to this problem is the
least squares solution(i.e. to have the hidden units span the L
principle components with the highest eigenvalues)
Cottrell et al. Found that the weights of 16 hidden units span the
space of the 13 first engenvectors of he covariance matrix of the
inputs.
when transformed to gray scale and graphically displayed, the
hidden unit receptive and projective fields looked “face-like” and
showed some similarity to the eigenvectors or eigenfaces
The hidden value of compression network is very
useful in face recognition!!
14. Further work
Improvement of the performance of face recognition
improvement of the compression network
If the compression network is trained with all face images,
what will be different from experiment 2 ?
The performance of compression network
The performance of recognition network
15. Reference
Dominque Valentin, Herve Abdi, Alice J. O’Toole, Garrison
W. Cottrell, “Connectionist Models of Face Processing: A
survey”, 1994