Using Evolutionary Algorithm to find image segmentation Yossef - - PowerPoint PPT Presentation

▶
using evolutionary algorithm to find image segmentation
SMART_READER_LITE
LIVE PREVIEW

Using Evolutionary Algorithm to find image segmentation Yossef - - PowerPoint PPT Presentation

Nov 19, 2022 11 likes •442 views

Using Evolutionary Algorithm to find image segmentation Yossef Kitrossky & Yoad Lewenberg Evolutionary Algorithm First Generation Population Evolutionary Algorithm First Generation Population Individual Individual B A Evolutionary

slide-1
SLIDE 1

Using Evolutionary Algorithm to find image segmentation

Yossef Kitrossky & Yoad Lewenberg

slide-2
SLIDE 2

Evolutionary Algorithm

First Generation Population

slide-3
SLIDE 3

Evolutionary Algorithm

First Generation Population

Individual

A

Individual

B

slide-4
SLIDE 4

Evolutionary Algorithm

First Generation Population

Individual

A

Individual

B

Individual

C

slide-5
SLIDE 5

Evolutionary Algorithm

First Generation Population

Individual

A

Individual

B

Individual

C

Individual

C’

slide-6
SLIDE 6

Evolutionary Algorithm

First Generation Population

Individual

A

Individual

B

Individual

C

Individual

C’ New Population

slide-7
SLIDE 7

First Generation

  • Random Matrix
  • Circles and rectangle
slide-8
SLIDE 8

First Generation

  • Random Matrix
  • Circles and rectangles
slide-9
SLIDE 9

First Generation

  • Random Matrix
  • Circles and rectangle
slide-10
SLIDE 10

First Generation

  • Random Matrix
  • Circles and rectangle

Mutation probability 0.02 Mutation probability 0.2

slide-11
SLIDE 11

First Generation

  • Random Matrix
  • Circles and rectangles

Mutation probability 0.02 Mutation probability 0. 2

slide-12
SLIDE 12

Evolution

 Reducing image resolution

128*128 16*16 32*32 64*64 8*8

slide-13
SLIDE 13

Evolution

slide-14
SLIDE 14

Evolution

20 generation of evaluation according to 8*8 resized image

slide-15
SLIDE 15

Evolution

40 generation of evaluation according to 16*16 resized image

slide-16
SLIDE 16

Evolution

80 generation of evaluation according to 32*32 resized image

slide-17
SLIDE 17

Evolution

100 generation of evaluation according to 64*64 resized image

slide-18
SLIDE 18

Evolution

160 generation of evaluation according to

  • riginal image
slide-19
SLIDE 19

Evolution

slide-20
SLIDE 20

Selection

 The best 10% individuals join to the next generation as

they are.

 For the last 90%:

 Randomly choose 4 individuals.  The best one chosen as parent A.  In the same way parent B is chosen.  The offspring of A and B, be a member of the next generation.

slide-21
SLIDE 21

Merge

 Randomly choose pivot  Randomly choose axis  With some probability mutate the result

slide-22
SLIDE 22

Merge

 Randomly choose pivot  Randomly choose axis  With some probability mutate the result

slide-23
SLIDE 23

Merge

 Randomly choose pivot  Randomly choose axis  With some probability mutate the result

Pivot = 54, y axis

slide-24
SLIDE 24

Mutation

Method 1

Flip random index

Method 2

Add circle

Add rectangle

Smooth

Segment expansion

slide-25
SLIDE 25

Mutation

Method 1

Flip random index

Method 2

Add circle

Add rectangle

Smooth

Segment expansion

slide-26
SLIDE 26

Mutation

Method 1

Flip random index

Method 2

Add circle

Add rectangle

Smooth

Segment expansion

slide-27
SLIDE 27

Mutation

Method 1

Flip random index

Method 2

Add circle

Add rectangle

Smooth

Segment expansion

slide-28
SLIDE 28

Mutation

Method 1

Flip random index

Method 2

Add circle

Add rectangle

Smooth

Segment expansion

slide-29
SLIDE 29

Fitness Function

 I=

2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2

slide-30
SLIDE 30

Fitness Function

 I=

2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 4 4 4 z 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2

slide-31
SLIDE 31

Fitness Function

 I=

2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2 4 4 4 4 4 2 2 2

slide-32
SLIDE 32

Fitness Function

 A=

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

slide-33
SLIDE 33

Fitness Function

 Low variance in each segment.  High derivative at boundary points

slide-34
SLIDE 34

Fitness

 A=

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

At boundary point by x axis, should receive high values

slide-35
SLIDE 35

Fitness

 Ix=

1 1 1 1 1 1 1 1 1 1

slide-36
SLIDE 36

Fitness

 A=

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

At boundary point by y axis, should receive high values

slide-37
SLIDE 37

Fitness Function

 Iy=

  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
  • 1
slide-38
SLIDE 38

Fitness

slide-39
SLIDE 39

Image with noise

slide-40
SLIDE 40

Image with noise

slide-41
SLIDE 41

Running time

 For n*n image:  Creating the initial population.  For every generation;  Ranking all the population  for every individual;  Pick parents  Merge  Mutate

 Total running time: -

slide-42
SLIDE 42

Running time

 For n*n image:  Creating the initial population.  For every generation;  Ranking all the population  for every individual;  Pick parents  Merge  Mutate

 Total running time:

slide-43
SLIDE 43

Running time

 For n*n image:  Creating the initial population.  For every generation;  Ranking all the population  for every individual;  Pick parents  Merge  Mutate

 Total running time: