University of Toronto - Winter 2018
Department of Computer Science

2523H: Object Modeling and Recognition: Shape Perception in Human and Computer Vision

Professor:                               Sven Dickinson
Lecture Time:                         Wednesday, 10am-12pm
Location:                                BA (Bahen) 7231 (Feb 14, 28 in BA5256)
Office Hours:                          Tuesday, 1-3pm, LP (Pratt) 283B (or by arrangement w/ instructor)
Recommended Preparation:   a prior course in human and/or computer vision would be helpful, but is not required.
 

Recent Announcements:
(see old announcements)

(March 26, 2018, 10:00am) Some reminders for the week:

Your last one-page project progress report is due in class this week. The final report is due two weeks from Wed (April 11), our last class.

This Wed (Mar 28), we have Han and John presenting on Deformable Models of Shape in 2-D.

Next Wed (April 4), we have Nancy and Hang presenting on Architectures for Shape Perception.

And the last lecture (April 11), we will have our "5 minute madness" presentations, with each student (or team) presenting a 5-minute overview of their project. I'll also leave some class time for you to enter course evaluations.

(February 12 2018, 3:00pm) Note that I will be out of town during reading week as well as Tuesday, Feb 27, and will not hold office hours on these days. I will, however, be reachable by email should you have any questions.

(February 7, 2018, 6:30pm) Note that I will be out of town during reading week and so there will be no office hours. I will, however, be reachable by email, in case you have any questions you need help with.

(February 5, 2018, 11:30am) Just a reminder that your project proposals are due this Wed in class.

(January 29, 2018, 6:00pm) Just a reminder that the presentation schedule has been posted.

(January 25, 2018, 12:30am) The deadline for project proposal submission has been extended by one week to February 7.

(January 22, 2018, 12:00pm) Just a reminder to come to class having read the two papers that will be presented this week. The papers can be found at the link to the syllabus at the bottom of this page.

(January 22, 2018, 12:00pm) Just a reminder to send me your top 5 paper presentation preferences by this Wednesday, before class, at the latest.

(January 21, 2018, 12:00pm) The slides I presented in the first class are available in the file Introduction.pdf in the Papers directory.

(January 16, 2018, 12:00pm) Reminder that tomorrow I repeat the first class for those that missed it last week (due to an admin error). Note that we have moved rooms, and are now in BA 7231.

(January 11, 2018, 3:00pm) On January 9 (the day before our first class), a one-week postponement (meant for a different course) was accidentally posted for our class on the department's graduate course schedule. As a result, a number of students missed the class, through no fault of their own. I will therefore be repeating the first class for those students during the next class (Jan 17); students that attended the first class can have next week off. The two students who volunteered to present the first two presentations will present instead on January 24 (one week delay). Finally, the course will be extended by one week to cover the syllabus (and week 12 presentations). My apologies for the inconvenience.

(January 11, 2018, 3:00pm) Lynda has secured us a larger and more convenient room for our course. Starting next week (Jan 17), the class will be held in Bahen (BA) 7231, except on Feb 14 and 28, when it will be held in BA 5256.

(January 4, 2018, 4:00pm) I'm looking for a volunteer to present the Witkin and Tenenbaum paper on January 17; I already have a volunteer for the Qi et al. paper.

(January 4, 2018, 4:00pm) The syllabus is now available; see the link below.

(January 4, 2018, 4:00pm) I'm searching for a larger room, as more people want to take the course. First class will still be in MS3290 to avoid confusion.


Description:

The shape of an object is a key feature for object categorization, and is invariant to changes in colour, texture, and illumination: a car with a leopard-skin paint job is not mistaken for a leopard, nor is an elephant painted with stripes mistaken for a zebra. How shape is perceived in the human visual system can both inform and inspire how shape can be represented and recovered in a computer vision system. This interdisciplinary graduate research seminar, suitable for students in human or computer vision, will examine shape perception from this dual perspective. Through a discussion of seminal research papers in both human and computer vision, the course will explore both classical and current models of shape perception, along with the challenges both communities face.

Evaluation:

The course will be run in a seminar-like format, in which 2 students will each present and lead the discussion on a research paper each week. The course grade will be based primarily on a class project (70%), which can take the form of a software system, a comprehensive survey paper, or a research exercise. The remainder of the grade will come from class participation (20%) and the paper presentation (10%). Students can work in groups of 1-2 on their course project, and are encouraged to meet with the instructor early in the semester to converge on a project that aligns well with their interests, their background, and, if applicable, their graduate research. Project proposals are due before class on January 31, 2018. Students will subsequently report their project progress in a set of biweekly progress reports and will hand in a final report at the last class.

References:

There will be no required textbook for the course; copies of published research papers will made available online or provided to the students.

Syllabus:

Click here for a syllabus of papers we'll be covering, and click here to see the current student presentation schedule.