This is presentation about pattern recognition

1. PATTERN
RECOGNITION
Talal A. Alsubaie
SFDA
1

2. OUTLINES
 What is a pattern?
 What is A pattern Class?
 What is pattern recognition?
 Human Perception
 Examples of applications
 The Statistical Way
 Human and Machine Perception
 Pattern Recognition
 Pattern Recognition Process
 Case Study
2

3. WHAT IS A PATTERN?
 A pattern is an abstract object, or a set of
measurements describing a physical object.
3

4. WHAT IS A PATTERN CLASS?
 A pattern class (or category) is a set of
patterns sharing common attributes.
 A collection of “similar” (not necessarily
identical) objects.
 During recognition given objects are assigned
to prescribed classes.
4

5. WHAT IS PATTERN RECOGNITION?
 Theory, Algorithms, Systems to put Patterns
into Categories
 Relate Perceived Pattern to Previously
Perceived Patterns
 Learn to distinguish patterns of interest from
their background
5

6. HUMAN PERCEPTION
 Humans have developed highly sophisticated
skills for sensing their environment and taking
actions according to what they observe, e.g.,
 Recognizing a face.
 Understanding spoken words.
 Reading handwriting.
 Distinguishing fresh food from its smell.
 We would like to give similar capabilities to
machines.
6

7. EXAMPLES OF APPLICATIONS
7

8. THE
STATISTICAL
WAY
8

9. GRID BY GRID COMPARISON
A A B
Grid by Grid
Comparison
9

10. GRID BY GRID COMPARISON
A A B
10
0 0 1 0
0 0 1 0
0 1 1 1
1 0 0 1
1 0 0 1
0 1 1 0
0 1 1 0
0 1 1 0
1 0 0 1
1 0 0 1
No of
Mismatch= 3

11. GRID BY GRID COMPARISON
A A B
Grid by Grid
Comparison
11

12. GRID BY GRID COMPARISON
A A B
12
0 0 1 0
0 0 1 0
0 1 1 1
1 0 0 1
1 0 0 1
1 1 1 0
0 1 0 1
0 1 1 1
0 1 0 1
1 1 1 0
No of
Mismatch= 9

13. PROBLEM WITH GRID BY GRID
COMPARISON
 Time to recognize a pattern - Proportional to
the number of stored patterns ( Too costly with
the increase of number of patterns stored )
13
Solution
Artificial
Intelligence
A-Z
A-Z a-z
a-z 0-9
0-9
*/-+1@#
*/-+1@#

14. HUMAN AND MACHINE PERCEPTION
 We are often influenced by the knowledge of how
patterns are modeled and recognized in nature when we
develop pattern recognition algorithms.
 Research on machine perception also helps us gain
deeper understanding and appreciation for pattern
recognition systems in nature.
 Yet, we also apply many techniques that are purely
numerical and do not have any correspondence in
natural systems.
14

15. PATTERN RECOGNITION
 Two Phase : Learning and Detection.
 Time to learn is higher.
 Driving a car
 Difficult to learn but once learnt it becomes
natural.
 Can use AI learning methodologies such as:
 Neural Network.
 Machine Learning.
15

16. LEARNING
 How can machine learn the rule from data?
 Supervised learning: a teacher provides a category label or
cost for each pattern in the training set.
 Unsupervised learning: the system forms clusters or natural
groupings of the input patterns.
16

17.  Classification (known categories)
 Clustering (creation of new categories)
CLASSIFICATION VS. CLUSTERING
17
Category “A”
Category “B”
Clustering
(Unsupervised Classification)
Classification
(Supervised Classification)

18. PATTERN RECOGNITION PROCESS
(CONT.)
18
Post- processing
Classification
Feature Extraction
Segmentation
Sensing
input
Decision

19. PATTERN RECOGNITION PROCESS
 Data acquisition and sensing:
 Measurements of physical variables.
 Important issues: bandwidth, resolution , etc.
 Pre-processing:
 Removal of noise in data.
 Isolation of patterns of interest from the background.
 Feature extraction:
 Finding a new representation in terms of features.
 Classification
 Using features and learned models to assign a pattern to a
category.
 Post-processing
 Evaluation of confidence in decisions. 19

20. CASE STUDY
 Fish Classification:
 Sea Bass / Salmon.
 Problem: Sorting incoming fish
on a conveyor belt according to
species.
 Assume that we have only two kinds of fish:
 Sea bass.
 Salmon.
20
Salmon
Sea-bass

21. CASE STUDY (CONT.)
 What can cause problems during sensing?
 Lighting conditions.
 Position of fish on the conveyor belt.
 Camera noise.
 etc…
 What are the steps in the process?
1. Capture image.
2. Isolate fish
3. Take measurements
4. Make decision
21

22. CASE STUDY (CONT.)
22
Classification
Feature Extraction
Pre-processing
“Sea Bass” “Salmon”

23. CASE STUDY (CONT.)
 Pre-Processing:
 Image enhancement
 Separating touching or occluding fish.
 Finding the boundary of the fish.
23

24. HOW TO SEPARATE
SEA BASS FROM SALMON?
 Possible features to be used:
 Length
 Lightness
 Width
 Number and shape of fins
 Position of the mouth
 Etc …
 Assume a fisherman told us that a “sea bass” is
generally longer than a “salmon”.
 Even though “sea bass” is longer than “salmon” on the
average, there are many examples of fish where this
observation does not hold. 24

25. HOW TO SEPARATE
SEA BASS FROM SALMON?
 To improve recognition, we might have to use
more than one feature at a time.
 Single features might not yield the best performance.
 Combinations of features might yield better performance.
25
1
2
x
x
￩
￪
￫
1
2
:
:
x lightness
x width

26. FEATURE SELECTION
26
“Good” features “Bad” features

27. DECISION BOUNDARY
27

28. DECISION BOUNDARY (CONT.)
28
More complex model result more complex boundary

29. DECISION BOUNDARY (CONT.)
29
Different criteria lead to different decision boundaries

30. DECISION BOUNDARY (CONT.)
 What if a customers find “Sea bass” in there
“Salmon” can?
 We should also consider costs of different
errors we make in our decisions.
30

31. DECISION BOUNDARY (CONT.)
 For example, if the fish packing company
knows that:
 Customers who buy salmon will object vigorously
if they see sea bass in their cans.
 Customers who buy sea bass will not be unhappy
if they occasionally see some expensive salmon in
their cans.
 How does this knowledge affect our decision?
31

32. CASE STUDY (CONT.)
Issues with feature extraction:
 Correlated features do not necessary improve
performance.
 It might be difficult to extract certain features.
 It might be computationally expensive to extract
many features.
 Missing Features.
 Domain Knowledge.
32

33. THE DESIGN CYCLE
33

34. DEMO
34

35. DEMO
 Online face detector demo:
 http://demo.pittpatt.com/index.php
35

36. 36

37. DEMO (CONT.)
 With my friend “Albert Einstein”
37

38. VIDEO DEMO
38

39. Q & A
39

40. THAN
K YOU
40