Course Philosophy

If presented in a purely linear fashion, computer vision can be a dry subject, in sharp contrast with the passion that faculty and students in the CPL and BORG labs feel for their research topic. In this course I introduce some novel ideas to make it more cutting-edge research oriented as well as geared towards the practice of computer vision, rather than merely theory. I hope you like it :-)

While I might have written a different book, I will nevertheless closely follow the Forsyth & Ponce book, as this enables you to absorb the material in at least two different ways: once in the lecture and once while reading the corresponding chapters in the book. The material will be reinforced through programming assignments throughout the semester.

Course schedule

The detailed course schedule can be found here. Following the book, the course is divided into 6 parts, I-VI. The chapters and corresponding lecture content is indicated for each week. The last part in the book, on applications, is optional reading. Three extra lectures on material that is only peripherally covered in the book are indicated in bold, and handouts (H) will be provided.

Assignments

There will be 3 MATLAB programming assignments that reinforce the material covered in the lectures and the book, numbered A1 through A3. A MATLAB tutorial will be scheduled for students that have not used it. Note that programming skills and linear algebra are both pre-requisites for the class, and you will need them.

You will have approx. 2 weeks for each assignment, with a preliminary deliverable due after the first week of each assignment, and are thus expected to start early. Assignments will need to be handed in in electronic format to the TA, in a format to be determined.

We plan to grade and return assignments promptly. As a result, I will require all assignments to be turned in on time. No late submissions will be accepted without prior permission of the instructor.

Final Project

The last 4 weeks in the course you will be asked to do a final project, which is of a slightly more prescribed form than in previous years. In particular, I will ask each one of you to implement an algorithm/approach described in a computer vision research paper that you like or find particularly interesting. The paper does not have to be recent: it might be fun to implement one of the seminal vision papers from the seventies or earlier! But it would be equally nice to see some cutting-edge papers implemented. Finally, graduate students in computer vision can take this opportunity to implement a competing or seminal approach to compare with their own.

To force you to think about it well in advance of the due-date, you need to hand in “notice of intent” on Oct 31, consisting of a hardcopy of the paper along with a one-page description of what exactly you intend to implement. We will provide an example well before that time.

In the final 3 lectures, everyone will get a chance to present their final project. The form and time allotment of the presentation will be very limited due to the high enrollment in the class.

The final project is due on Dec 2 in a format to be determined, along with powerpoint slides (yes, powerpoint!). No late submissions will be accepted due to the logistics associated with the in-class presentations.

Examinations

Both a midterm and a final exam will be given.  The midterm will cover material presented in the first 2/5 of the course, while the final exam will cover the remaining 3/5 of the course material.