Kursvorlesung/Weiterführende Informatik-Veranstaltung:
Fundamentals of Pattern Recognition
The Chair for Pattern Recognition and Image Processing offers
yearly in the fall term a foundation course in pattern recognition.
| Description |
| The course deals with basic
methods used in pattern recognition. First of all, its relation with the more general estimation theory and the most important applications are discussed. Then, the basics of pattern recognition are introduced, including the concept of equivalence classes, position invariant feature extraction and the characteristics of completeness and separability in invariance theory. In the following chapter, fast non-linear algorithms for translation invariant classification for grayscale images are dealt with. Later, similarity and affine invariant features for contour images are derived, and the mapping properties and computational complexity of these so-called Fourier descriptors are discussed. The later part of the course deals with classifier design. First of all, optimal classifiers in the stochastic sense are introduced. In a later part, the solution for the basis of non-linear regression as well as learning theory is thoroughly dealt with. This includes the optimum polynomial classifier as well as the learning theory with neural networks operating with the help of backpropagation. In the end the so-called support vector machines are introduced, which is a new statistical learning approach which learns from the given training samples and has achieved excellent results in different applications. |
| Key data | |
| Type: | Kursvorlesung |
| Lecturer: | Prof. Dr. Ing. H. Burkhardt |
| Time/Place: | Wed 11-13, Fr 11-12, Building 101 HS 26 |
| Exercices: | see tutorials' wiki. The Wiki pages are readable after registration and login in the Wiki. |
| Beginning: | Fr, 23/10/2009 |
| Credit Points: | 6 Credit Points |
| Winter Exam: | Written exam: 25.03.2010, 14.00 o'clock, Building 101 026/036 |
| Lecture Slides: | slides from WS09/10 |
| Exam preparation |
| Date and time of additional tutorials for questions concerning the exam and additional exercises can be found in the Übungswiki. You have to register in the Wiki to have read access. If you have problems with this, please ask us. |
| Admission requirements to take part in the exam |
| If you achieve 60% of the maximum number of points that can be gained by handing in exercices and present one exercice in front of your exercice group, you get the so called Übungsschein which states that you have taken part at the exercices. This Übungsschein is a prerequisite to be allowed to take part in the exam! |
| Contents |
| 1. Introduction and Application Areas |
| 2. Basics of Pattern Recognition (equivalence classes, position invariant feature extraction) |
| 3. Position Invariant Grayscale Image Detection (the CT class, parallel implementation, extension to 2-D case, effects of systematic and stochastic noise, clustering properties) |
| 4. Position Invariant Contour Image
Detection (contour extraction, Fourier analysis, Fourier descriptors for the equivalence class of similarity and affine transformed patterns). |
| 5. General Approaches for Computing Invariants |
| 6. Feature Reduction, Feature Selection |
| 7. The Optimum Classifier, MAP- and MLE-criteria, Metrics |
| 8. Learning Strategies with Neural Networks |
| 9. The Polynomial Classifier |
| 10. Support Vector Machines |
| Excercises |
| Tutorials are aimed at consolidating the theoretical
insight gained in the lectures. You have to hand in weekly sheets of exercices. These are corrected and discussed in the following week. Exercice sheets can be found online. link to the tutorials' wiki. The Wiki pages are readable after registration and login in the Wiki. |
| Lecture Slides |
| Regularly updated lecture slides from the current semester can be found within the tutorials' wiki here. The Wiki pages are readable after registration and login in the Wiki. |
| Further Material |
| Slides in English (created in WS04/05, updated from time to time) |
| Video lecture recordings (recorded in WS02/03) |
| Exercices and solutions from WS05/06 and slides from WS05/06 |
Albert-Ludwigs-University Freiburg,
Chair for Pattern Recognition and Image Processing