Course Description and Overview
This seminar will survey the state of the art in computer vision through readings of original papers and implementation of classic algorithms. Beginning with the basics of color theory and camera models, the course will look at processing steps in a typical image pipeline. After considering low-level feature extraction such as edge detection, optical flow, and stereo correspondence, the course will take up higher-level issues such as object segmentation and tracking, structure from motion, and image comparison and retrieval. Prerequisites: CSC 112, MTH 153.
The course will cover some subset of the following list of topics.
Goals
You should come out of this course with a sense of the fun, possibilities, and challenges inherent in computer vision. You will learn to read and analyze a computer science research paper, and present its significant results. You will gain some familiarity with Matlab. Finally, you will see the sorts of problems currently under study by researchers in computer vision.
Course Materials
- Required
- Machine Vision, by Ramesh Jain, Rangachar Kasturi, and Brian G. Schunck. (McGraw-Hill) -- This has good coverage of basic image processing techniques.
- Recommended
- Computer Vision: A Modern Approach, by David A. Forsyth and Jean Ponce. (Prentice Hall; ISBN: 0130851981) -- This is a very complete text, but is written at an advanced graduate student level. It has more mathematical detail, and may be useful as a reference. I have a copy that may be loaned out for short periods.
- MATLAB student edition (computer software) -- this will allow you to install MATLAB on your personal computer instead of working in the lab.
I have additional books on computer vision that may be loaned out for short periods.
Requirements
This course will assume proficiency in programming, but none specifically with the MATLAB environment.
Expected work includes:
- Weekly assignments. In addition to readings, there will be exercises assigned each week. These will generally be short and to the point, designed to provide hands-on interaction with the topic of the week. They may be programming assignments or paper-and-pencil.
- In-class presentations of recently published research in computer vision. Each week we will be reading scholarly research papers in computer vision. Students will be expected to choose three to four papers over the course of the semester on which to prepare a presentation and discussion for the class as a whole. All students will be expected to have read the paper in advance so that they can participate in the discussion.
- Final programming project with in-class presentation. In lieu of a final, each student will complete a significant implementation project and present the results to the class. Typically, this will involve implementing a vision algorithm described in the scientific literature.
Collaboration policy: Because the details of each student's project will differ, some consultation on technical aspects is permitted. Students may discuss homework orally but may not share written code. In general, one student may not create content for submission under another student's name, and any work submitted must accurately reflect the understanding of the student who submitted it. Abuses of this policy will result in a referral to the Honor Board.
Grading
- Weekly Assignments 30%
- Paper Presentations 30%.
- Final Project and Presentation 20%.
- Class Participation 20%
The numbers above should be taken as a rough indication of the relative importance of various components. It is expected that every student will contribute regularly to class discussions. Therefore the participation component will typically act as a potential modifier on the overall grade, only raising or lowering it in cases where a student's performance differs significantly from the norm.
Assignments will not normally be accepted late, except at the sole discretion of the instructor, and may be assessed a grade penalty. If a student is unprepared to present a paper for which they have signed up, there may be no opportunity to reschedule, resulting in a zero grade for the assignment.
Students are encouraged to read the instructor's policy on grade averaging.
Course Details
- Professor
- Nick Howe
McConnell 215
585-3878
nhowe@cs.smith.edu
Office hours - Meeting Times
- 10:30-11:50 AM Tuesdays & Thursdays
- Room
- Burton B17