TU Delft
print this page print this page     
2013/2014 Electrical Engineering, Mathematics and Computer Science Master Computer Science
Computer Vision
Responsible Instructor
Name E-mail
Dr. E.A. Hendriks    E.A.Hendriks@tudelft.nl
Dr. L.J.P. van der Maaten    L.J.P.vanderMaaten@tudelft.nl
Contact Hours / Week x/x/x/x
0/0/0/2 en 4 uur werkcollege in 4e Q
Education Period
Start Education
Exam Period
Course Language
Expected prior knowledge
You are expected to have a working understanding of linear algebra, and of probability and statistics. Knowledge about pattern recognition and/or machine learning is preferred.
Course Contents
The central theme of the computer vision course is the automatic analysis and interpretation of images and videos using computer algorithms. The course explores a range of techniques for image segmentation, object detection, object tracking, object recognition, image alignment, and scene understanding.

The course consists of seven two-hour lectures, and seven four-hour hands-on sessions. In each lecture, techniques for a particular problem are introduced, and some applications of these techniques are described. During the hands-on sessions, you will make exercises that center around these applications. These exercises are designed to improve your understanding of techniques for computer vision. In the exercises, you will implement (parts of) systems for applications such as augmented reality, constructing panoramic pictures from collections of photographs, recognizing the identity of faces, and gesture recognition. Active participation in the hands-on sessions is mandatory; the solutions to assignments must be shown to the instructor or one of the TAs.
Study Goals
After successfully completing this course:

- You are able to explain and implement various techniques for feature point detection, and can explain the type of feature points these detectors identify.
- You are able to explain and implement techniques for feature point description and feature point matching. You are able to use these techniques in applications such as object detection and image stitching.
- You are able to explain and implement techniques for image stitching. The student understands the key problems in developing image-stitching algorithms (such as alignment and parallax removal).
- You are able to explain and implement techniques for shape analysis.
- You are able to explain and implement techniques for face detection and face recognition.
- You are able to explain and implement techniques for object recognition and scene understanding.
- You are able to explain and implement basic techniques for feature tracking, in particular, Kanade-Lucas-Tomasi tracking and particle filter tracking.
- You are able to explain Markov Random Field models, and is able to use such models in problems such as image denoising and inpainting.
- You are able to develop and explain computer vision systems for real-world applications. In particular, you are able to select computer vision techniques that are to solve a specific image analysis or image understanding problem, to motivate this selection, and to combine the selected techniques into a working computer vision system.
Education Method
- Section 1, 2, 3, 4, 6, 9, 10.5, and 14; Appendix B of “Computer Vision, Algorithms and Applications”, R. Szeliski, Springer, 2011, ISBN 978-1-84882-934-3. (This book is freely available online.)
- "Shape Matching and Object Recognition Using Shape Contexts", S. Belongie, J. Malik, and J. Puzicha. IEEE Transactions on Pattern Analysis and Machine Intelligence 24(24): 509–521, 2002.
- “Discriminative random fields”, S. Kumar and M. Hebert, International Journal of Computer Vision 68(2):179–202.
- “Fields of Experts”, S. Roth and M.J. Black, International Journal of Computer Vision 82(2):205–229, 2009.
- Section 1 and 2 of "Lucas-Kanade 20 Years On: A Unifying Framework", S. Baker and I. Matthews, International Journal of Computer Vision 56(3):221–255, 2004.
- "CONDENSATION — Conditional Density Propagation for Visual Tracking", M. Isard and A. Blake, International Journal of Computer Vision 29(1):5-28, 1998.
Prerequisite courses (mandatory): Image or signal processing (TI2710-A or TI2710-C or EE2521); Linear algebra (WI1200TI-A or WI1200TI-B or WI1142TN); Probability and statistics (WI2211TI or WI1120EE or WI1102CT or WI1321TB or WI2013wbmt). If you have take comparable courses at other universities, this is also fine. Please contact the instructor when in doubt!

Prerequisite courses (preferred): Pattern recognition (IN4085).
The assessment for this course consist of two main parts:

1) You will develop a computer vision system for an application of your choice, and will write a small report with a description of your system. Your grade for the project forms 50% of the final grade for the course.

2) A written final exam will determine the remaining 50% of your final grade. The final exam covers: (1) all content covered in the lectures and (2) all material listed under “Course material”. The final exam is an open-book exam: all slides and course material can be used during the exam.