Probabilistic Graphical Models for Image Analysis

Prof. Joachim Buhmann, Dr. Cheng Soon Ong - Autumn Semester 2009

Course Description

This course will focus on inference with statistical models for image analysis. We use a framework called probabilistic graphical models which include Bayesian Networks and Markov Random Fields. We apply the approach to traditional vision problems such as shape from shading and image segmentation, as well as recent problems such as object recognition.

Time and Place

Lectures	Wed 8-10, CAB H52
Exercises	Wed 10-11 CAB H52

Exercises

Exercise problems will include theoretical and programming problems. Programming will be done in Matlab. Detailed exemplary solutions will be distributed for all exercises.

Performance Assessment

To obtain a Testat (course attendance confirmation), you will be required to attend exercise classes, turn in problem solutions and achieve 50% of all possible points therein.

Please note: We cannot tell you whether you need a Testat or not, only what the requirements are in order to get one for this course. Please consult with the student's administration in your departement. For computer science students, a Testat is usually not required.

Exam

30 Minute oral exam in English.

Syllabus

Week	Lecture Topics	Lecture Slides	Exercise Sheets	Exercise Solutions	Additional Material
Sep 16th	Introduction	Lecture 1 notes to lecture 1 notes on probability	Series 01 no exercise session!	Solution 01
Sep 23rd	Logistic Regression	Lecture 2 notes to lecture 2	Series 02	Solution 02	Handout exercise 1 LR code template LR solution
Sep 30th	Clustering	Lecture 3 notes to lecture 3	Series 03	Solution 03	Handout exercise 2 Segmentation template Segmentation solution
Oct 7th	Graphical Models	Lecture 4 notes to lecture 4	Series 04	Solution 04	Chapter 8 of Bishop available from his book website Handout exercise 3 Ancestral template Ancestral solution
Oct 14th	Sum Product algorithm	Lecture 5 Refer to Chapter 8 of Bishop for details.	Series 05	Solution 05	Chapter 16 of Mackay. Handout exercise 4
Oct 21st	MRF	Lecture 6 notes to lecture 6	Series 06	Solution 06	MRF code Handout exercise 5
Oct 28th	Graph-Cut	Lecture 7	Series 07	Solution 07	Handout exercise 6 LP inference code (updated)
Nov 4th	Sampling	Lecture 8 Refer to Chapter 29 of Mackay for details.	Series 08	Solution 08	Handout exercise 7 Gibbs framework Gibbs solution Error in proof of detailed balance in the printed copy of Bishop's book.
Nov 11th	Variational Methods	Lecture 9 Notes to lecture 9 Notes about exponential families	Series 09	Solution 09	Handout exercise 8
Nov 18th	Support Vector Machines	Lecture 10 Tutorial paper	Series 10	Solution 10	Handout exercise 9
Nov 25th	Conditional Random Fields	Lecture 11 Notes to lecture 11	Series 11	Solution 11	Handout exercise 10
Dec 2nd	Structured SVM	Lecture 12 Tutorial at ICCV'09 Kernel Methods in Computer Vision	Series 12	Solution 12
Dec 9th	Modelling	Lecture 13
Dec 16th	Multiple Kernel Learning	Lecture 14

Slides contain copyrighted material from various sources and are intended for use in the course only.

Resources

References

C. Bishop. Pattern Recognition and Machine Learning. Springer 2007.
This is an excellent introduction to machine learning that covers most topics which will be treated in the lecture. Contains lots of exercises, some with exemplary solutions. Available from ETH-HDB and ETH-INFK libraries.

D. Koller and N. Friedman. Probabilistic Graphical Models: Principles and Techniques. The MIT Press 2009.
Very recent book that covers Bayesian networks and undirected graphical models in great detail.

David J.C. Mackay. Information Theory, Inference and Learning Algorithms. Cambridge University Press, 2003.
Available for free from here.

Rafael C. Gonzalez and Richard E. Woods. Digital Image Processing. Prentice Hall, 3rd edition, 2007.

Matlab

The official Matlab documentation is available online at the Mathworks website (also in printable form). If you have trouble accessing Matlab's built-in help function, you can use the online function reference on that page or use the command-line version (type help <function> at the prompt). There are several primers and tutorials on the web, a later edition of this one became the book Matlab Primer by T. Davis and K. Sigmon, CRC Press, 2005.

Contact

Instructors: Prof. J. M. Buhmann, Dr. Cheng Soon Ong
Assistant: Patrick Pletscher