bild
Skolan för
elektroteknik
och datavetenskap

Image Based Recognition and Classification (bik12)

This course will focus on methods for recognition and classification of image content such as face detection and written text. There are two main strands to this course. The first deals with introducing the students to image processing techniques to extract information, feature extraction, from images that can be used for the classification of its content.

While the second and more substantial part gives a thorough overview at a practical and theoretical level of methods to perform the classification. Most of these methods rely on machine learning techniques and probabilistic reasoning.

Throughout algorithms introduced in the course will be tested and/or implemented by the students on real world images.

Course Aims

The goal of this course is to
  • Provide the students with the technical knowledge to be able to implement and assess standard machine learning techniques when applied to solving image recognition problems.
  • Give the students the practical experience and confidence to exploit their new technical know how on concrete problems.
  • Encourage students to be able to reason about which image recognition problems can be solved using machine learning approaches.

Basic Information

This is a 6 credit course and run by the school of Computer Science and Communication (CSC) at KTH. The people organizing the course are:

Course Leader: Josephine Sullivan (sullivan@nada.kth.se)
Teaching assistant: Miroslav Kobetski (kobetski@kth.se)

Registration

Only undergraduates registered in Ladok are eligible to take the course. You are supposed to first apply/choose this course as a part of your optional courses. Phd students will need a special form signed by their supervisor and the dean from their home institution. More info can be found here.

Everyone attending the course has to also register with CSC's course administration system res. This is done by using the following command

res checkin bik12

on one of the CSCs unix machines.

Before the first scheduled lab moment, you should also join the course with the CSC's course administration program course.

course join bik12

This command sets up your environment setting such as for example access to some of the module files needed during the course. These will be loaded automatically each time you log in. In addition, it controls with each login if a lecturer has sent any new messages to course participants.

After finishing the course, you should do

course leave bik12

to remove the effects of course join.

Course Structure at a Glance

Instruction
11 x 2 hour Lectures
2 x 2 hour Exercise classes
4 x 3 hour Lab sessions
Course work
9 Homework assignments
1 Lab Project
1 Final written exam

Course Reading Material

The lecture notes are self-contained. However, the following books cover the topic of classification and can be used for supplementary reading.

  • This book is the one closest to the lecture notes: Pattern Classification, Richard O. Duda, Peter E. Hart and David G. Stork, Wiley Interscience.

  • However, a cheaper alternative is: Pattern Recognition and Machine learning, Christopher M. Bishop, Springer.

  • Much of the material covered in this course is described in more detail in the book Computer Vision: Models, Learning, and Inference by Simon Prince. It is not for sale yet. However, an electronic version is available for downloading from the linked site.

  • Another interesing book available on the web is The Elements of Statistical Learning by Trevor Hastie, Robert Tibshirani and Jerome Friedman. This book focuses on machine learning as opposed to machine learning for computer vision. It is quite a long book but many parts of it are quite readable.
Copyright © Sidansvarig: Josephine Sullivan <sullivan@nada.kth.se>
Uppdaterad 2012-02-17