Face Recognition Using Artificial Neural Network Group-based Adaptive Tolerance Trees

1
Face Recognition Using Artificial Neural Network
Group-based Adaptive Tolerance Trees

• By Ming Zhang , John Fulcher

2
ReviesGroup Theory
3
ReviewNN Group Theory
4
Translation-invariant face recognition system
5
NN group-based tree node
6
NN group-based tree node
7
The Features of OR NN group
8
Advantage

• No single neural network is capable of
  approximating such a function comprising three
  peaks and nonsmooth , noncontinuous points

9
The Features of AND NN group
10
Advantage

• No single neural network is capable of
  approximating such a function comprising
  sole-peak and nonsmooth , and /or noncontinuous
  points

11

• Both OR and AND NN groups were used as the nodes
  for GAT tree, resulting in more accurate and
  efficient face recognition .

12
NN group-base adaptive tolerance tree.
13
GAT tree model
14
OPT

• Translation operator that can translate facial
  image MI(I,j) into a center face, left face,
  right face and so on --- shifts and rotates the
  facial image in two dimensions.(but only during
  training )

15
OPA

• Adaptive node operator set that adds adaptive
  connection and grows nodes in the GAT tree if the
  parent node output is within tolerance.

16
OPN

• Node operator set, which is a complex pattern
  classifier.

17
OPP

• Path operator set which sets the parent node
  output to the input of the child node.

18
OPL

• Label leaf operator set which indicates the
  labeled person has been recognized.

19
Face perspective classification using GAT tree
20
Face perspective classification results
21
Face perspective classification results
22
Face perspective classification results
23
Front glasses and beard face classification using
GAT tree.
24
Front beard face and Front glasses face
classification
25
Front beard face and Front glasses face
classification
26
GAT tree for front face recognition
27
Conclusion