Dr. Ghassabi [email protected]Tehran Markaz University Spring 2015 Statistical Pattern Recognition Session 1 (Introduction) 1 Course Link: cipcv.ir/lessons
40
Embed
Dr. Ghassabi [email protected] Tehran Markaz University Spring 2015 [email protected] Statistical Pattern Recognition Session 1 (Introduction)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Slide 1
Dr. Ghassabi [email protected] Tehran Markaz University
Spring 2015 [email protected] Statistical Pattern Recognition
Session 1 (Introduction) 1 Course Link: cipcv.ir/lessons
Slide 2
Grades Home work 20% Exam- 40% Presentation and Participation-
40% If you write a good paper, your grade will become 20 2
Slide 3
Homework TA: Ali Rahmani, Email:
[email protected]@gmail.com Format
(email(subject)& exercise(name)): SPR-NAME-Family-Exe1
Deadline: 1 week Group works: More than 1 week 3
Slide 4
Paper Presenting a paper (Survey or Research paper) An overview
of Statistical pattern recognition techniques for speaker
verification An overview of Statistical pattern recognition
techniques for speaker verification Research paper (2010,,2014)
Comparing results of several papers (an evaluation study) Writing a
research paper 4
http://www.matlabsite.com/1127/mvpacw9210-choice-of-topic-and-thesis-writing.html
http://www.matlabsite.com/1748/fvacw9304-practical-thesis-preparation-and-academic-
paper-publication.html
Slide 5
Presentation Example:
http://research.cs.queensu.ca/home/xiao/pr.html
http://research.cs.queensu.ca/home/xiao/pr.html Scope of subjects:
Image Processing and Computer Vision: Face Recognition, Object
Recognition, Document analysis, Biometric, remote sensing Speech
Recognition Datamining and machine learning 5
Slide 6
Computer Vision Robotics Neuroscience Graphics Computational
Photography Machine Learning Medical Imaging Human Computer
Interaction Optics Image Processing Feature Matching
Recognition
Slide 7
How To find a paper? IEEE, Elsevier, springer
http://cipcv.ir/related-magazine 7
Slide 8
How to find a Project Subject? ICPR (International Conference
on Pattern Recognition) ICPR 2013, 2012, 2010
http://www.icpr2012.org/overview.html 8
Slide 9
How to find a Project Subject? See Scope In Call Papers
http://www.icpr2012.org/cfp.html 9
Slide 10
How to find a Project Subject? http://cipcv.ir/related-magazine
Journals (IEEE, Elsevier, Springer) IEEE Transactions on pattern
analysis and machine intelligence Pattern Recognition Pattern
Recognition Letter 10
Slide 11
Goals Course Website:
http://cipcv.ir/lessonshttp://cipcv.ir/lessons To make the graduate
students acquainted with the fundamental concepts of statistical
pattern recognition and its applications in computational
intelligence Image Processing Computer Vision 11
Slide 12
Links Courses on Statistical Pattern Recognition SPR-Sharif
University A Statistical Learning/Pattern Recognition Glossary A
Statistical Learning/Pattern Recognition Glossary Tutorials on
Topics in Statistical Pattern Recognition Tutorials on Topics in
Statistical Pattern Recognition 12
Slide 13
Books Main books: Pattern Recognition books website (By
Theodoridis and Koutroumbas) Pattern Recognition books website (By
Theodoridis and Koutroumbas) Pattern classification (By duda) An
introduction to statistical pattern recognition(By fukunaga) Other
books: Fuzzy Models and Algorithms for Pattern Recognition and
Image Processing( by James C. Bezdek, James Keller, Raghu
Krisnapuram,Nikhil R. Pal) Fuzzy Models and Algorithms for Pattern
Recognition and Image Processing( by James C. BezdekJames
KellerRaghu KrisnapuramNikhil R. Pal Pattern Recognition and
Machine Learning books website (By Bishop) Statistical Data Mining
Tutorials (By Andrew Moore) Probabilistic Graphical Methods books
website (By Koller and Friedman) The Elements of Statistical
Learning books website (By Hastie, Tibshirani and Friedman) The
Elements of Statistical Learning books website (By Hastie,
Tibshirani and Friedman) 13
Slide 14
Papers A. K. Jain, R. P. W. Duin, J. Mao, "Statistical Pattern
Recognition: A Review" (local copy), IEEE Trans. on Pattern
Analysis and Machine Intelligence, 22(1):4-37, January
2000."Statistical Pattern Recognition: A Review"local copy Chi Hau
Chen, Pei-Gee Peter Ho, Statistical pattern recognition in remote
sensing, Pattern Recognition 41 (2008) 2731 2741 14
Slide 15
Matlab Toolboxes Statistical Pattern Recognition Toolbox
Statistical Pattern Recognition Toolbox PRTools: The Matlab Toolbox
for Pattern Recognition PRTools: The Matlab Toolbox for Pattern
Recognition Bayes Net Toolbox Bayes Net Toolbox HMM Toolbox HMM
Toolbox LIBSVM A Library for SVM LIBSVM A Library for SVM 15
Slide 16
Outline of PR course Introduction to PR and its applications
Bayes Theory Neural Networks (Perceptron, Multi layer perceptron,
RBF) Non-parametric methods (Decision Tree) SVM Clustering methods
Feature Extraction and Selection 16
Slide 17
Outline of PR course Introduction to PR and its applications
Bayes Theory Neural Networks (Perceptron, Multi layer perceptron,
RBF) Non-parametric methods (Decision Tree) SVM Clustering methods
Feature Extraction and Selection 17
Slide 18
What is Pattern? Opposite to chaos; it is an entity, object,
process or event, vaguely defined, that can be given a name or
label. Examples of Patterns: A fingerprint image A handwritten
cursive word A human face A speech signal Texture Etc. 18
Slide 19
What is PR? Systems to put patterns into pre-specified
categories Relate Perceived Patterns to previously Perceived
Patterns Classification What is a pattern? A fingerprint images, A
handwritten cursive word, A human face, A speech signal What kinds
of category we have? 19
Slide 20
An Example N pattern categories (classes) Sky, buildings,
mountain, trees, Common attributes (features) Color Contrast
Texture Goal Observing some labeled pixels, We wish to assign a
label to each new (unlabeled) pixel. 20
Slide 21
Application of PR Machine Vision Character Recognition Computer
aided diagnosis Speech Recognition Datamining and knowledge
Discovery Others: fingerprint identification, signature
authentication, text retrieval, and face and gesture recognition.
21
Slide 22
Application of PR 22
Slide 23
Exercise 1 What is the difference between PR and Machine
Learning (ML)?
http://www.icpr2014.org/tutorialpages/philosophicalaspects
http://www.icpr2014.org/tutorialpages/philosophicalaspects
http://stats.stackexchange.com/questions/5026/what-is-the-difference-between-data-
mining-statistics-machine-learning-and-ai
http://stats.stackexchange.com/questions/5026/what-is-the-difference-between-data-
mining-statistics-machine-learning-and-ai
http://www.37steps.com/638/machine-learning-and-pattern-recognition/
http://www.37steps.com/638/machine-learning-and-pattern-recognition/
http://www.quora.com/What-is-the-difference-between-pattern-recognition-and-
machine-learning
http://www.quora.com/What-is-the-difference-between-pattern-recognition-and-
machine-learning
http://www.inf.ed.ac.uk/teaching/courses/mlpr/lectures/mlpr-prelim.pdf
http://www.inf.ed.ac.uk/teaching/courses/mlpr/lectures/mlpr-prelim.pdf
23
Slide 24
Components of a PR system Patterns: Images Classes: (A, B) 24
Examples of image regions corresponding to (a) class A and (b)
class B. a benign lesionmalignant (cancer),
Slide 25
Components of a PR system What measurable quantities make these
two regions distinct from each other? 25 Plot of the mean value
versus the standard deviation for a number of different images
originating from class A (o) and class B (+). a straight line
separates the two classes. the unknown pattern is more likely to
belong to classA than class B
Slide 26
Components of a PR system Which features are discriminative? 26
Defective Cookies Non-Defective Cookies Are the mean value versus
the standard deviation suitable features to separate defective and
non-defective cookies?
Slide 27
Components of a PR system Which features are suitable to
recognize vessel and non-vessel features? 27
http://cvip.computing.dundee.ac.uk/papers/Lupascu2009.pdf
Slide 28
improfile: : : imshow(retina1.tif);improfile; 28
Slide 29
Components of a PR system the mean value and the standard
deviation are known as features l features x i, i =1, 2,..., l are
used, and they form the feature vector x [x1, x2,..., xl ] T Each
of the feature vectors identifies uniquely a single pattern
(object). The straight line is known as the decision line, and it
constitutes the classifier whose role is to divide the feature
space into regions that correspond to either class A or class B.
29
Slide 30
Components of a PR system How are the features generated?
Feature Extraction What is the best number l of features to use?
Feature Selection Having adopted the appropriate, for the specific
task, features, how does one design the classifier? Classifier
stage design Once the classifier has been designed, how can one
assess the performance of the designed classifier? That is, what is
the classification error rate? system evaluation stage. 30
Slide 31
Components of a PR system The design of a pattern recognition
system essentially involves the following three aspects: data
acquisition data representation decision making 31 The basic stages
involved in the design of a classification system.
Slide 32
Components of a PR system Categories (Classes): Supervised
Classification Discriminant Analysis Unsupervised Classification
Clustering 32
Slide 33
Design cycle Data collection Feature Choice Model Choice
Training Evaluation Computational Complexity 33
Slide 34
Image Categorization Training Labels Training Images Classifier
Training Training Image Features Testing Test Image Trained
Classifier Outdoor Prediction
Slide 35
Learning a classifier Given some set of features with
corresponding labels, learn a function to predict the labels from
the features xx x x x x x x o o o o o x2 x1
Slide 36
Formulation: binary classification Formulation +1 x1x1 x2x2
x3x3 xNxN x N+1 x N+2 x N+M ??? Training data: each image patch is
labeled as containing the object or background Test data Features x
= Labels y = Where belongs to some family of functions
Classification function Minimize misclassification error (Not that
simple: we need some guarantees that there will be
generalization)
Slide 37
Many classifiers to choose from SVM Neural networks Nave Bayes
Bayesian network Decision Trees K-nearest neighbor Etc. Which is
the best one?
Slide 38
PR approaches Statistical PR Based on underlying statistical
properties/model of patterns and pattern classes. Use numerical
features for distinguishing between classes. Bayesian Methods
Neural Networks Decision Trees Support Vector Machines Etc.
Structural (or syntactic) PR Based on explicit or implicit
representation of a classs structure Pattern classes represented by
means of formal structures as grammers, automata, strings, graphs,
trees, etc. Reference: Syntactic and structural pattern
recognition: theory and applications, By Horst Bunke, Alberto
Sanfeliu 38
Slide 39
PR approaches 39
Slide 40
Next Session Classifiers Based on Bayes Decision Theory 40