Computer Vision
Fall
2009
Tues/Thurs 11:00 am – 12:15 pm
CPE 2.206
CS 378, Unique # 54875
Instructor: Kristen
Grauman
Email: grauman @ cs
Office location: CSA 114
Office hours: Wednesdays 5-6 pm, and by
appointment.
TA: Jaechul Kim
Email: jaechul @ cs
Office hours: Mondays 3-4 pm, and Wednesdays
3-4 pm, in TAY basement computer lab.
TA: Yong Jae Lee (for office hours only)
Office hours: Tuesdays 4-5 pm, and Thursdays
4-5 pm, in TAY basement computer lab.
Please come to any of our office hours for
questions about assignments or lectures.
Questions via email about assignments should be
sent to: cv-fall2009 @ cs , with “378” in the
beginning of the subject line.
This will help ensure a timely response from
the instructor or TAs.
Schedule eGradebook Blackboard
Announcements
Answers for Section 1 of Pset 5 are available
here.
Final exam is Monday 12/14, 2-5 pm in
JGB 2.218. Example exams handed out in
class on 12/1.
Check out class mosaic results here.
Check out class seam carving results here.
Use the course schedule
for all reading assignments, deadlines, lecture notes, etc. Lecture
slides are linked from the 2nd to last column.
Grades
and late-days posted on eGradebook.
Overview
Billions
of images are hosted publicly on the web---how can you find one that “looks
like” some image you are interested in?
Could we interact with a computer in richer ways than a keyboard and
mouse, perhaps with natural gestures or simply facial expressions? How can a robot identify objects in complex
environments, or navigate uncharted territory?
How can a video camera in the operating room help a surgeon plan a
procedure more safely, or assist a radiologist in more efficiently detecting a
tumor? Given some video sequence of a
scene, can we synthesize new virtual views from arbitrary viewpoints that make
a viewer feel as if they are in the movie?
Computer
vision is at the heart of many such questions: the goal is to develop methods
that enable a machine to “understand” or analyze images and videos. In this introductory computer vision course,
we will explore various fundamental topics in the area, including image
formation, feature detection, segmentation, multiple view geometry, recognition
and learning, and motion and tracking.
An outline of the syllabus is here.
This
course is intended for upper-level undergraduate students.
Prerequisites
Basic
knowledge of probability and linear algebra; data structures, algorithms;
programming experience.
Previous
experience with image processing or machine learning will be useful but is not
required. Problem sets will include
Matlab programming problems. There will
be a Matlab tutorial in class on Thursday, September 3, and a warm-up
assignment to get familiar with basic Matlab commands. We will recommend useful functions to check
out per assignment. Students are
expected to practice and pick up Matlab on their own in order to complete the
assignments. The instructor and TAs are
happy to help with Matlab issues during office hours.
If
you are unsure if your background is a good match for this course, please come
talk to me.
Problem sets: Problem
sets will be given approximately every two weeks, and will involve a
combination of concept questions and programming problems. The programming problems will provide
hands-on experience working with techniques covered in or related to the
lectures. All code and written responses
must be completed individually. These
problem sets will take time to complete; please start early!
Small assignments: In addition to the problem sets
above, we may have a few very short assignments due within 3-7 days of their assignment.
Due dates: All
problem sets are to be submitted by 11:59 PM on the day they are due unless
otherwise noted on the assignment itself.
Deadlines are firm, and we will not make exceptions. We will use the “turnin” program timestamp
to determine time of submission.
Anything from 1 minute to 24 hours is one day late (i.e., a timestamp of
12:00 AM or later is one day late).
The
instructions in each problem set will designate which parts to submit
electronically and which if any to submit via hardcopy. Any hardcopy portions should be submitted in
class or left in Jaechul’s office, CSA 1.134.
Create a pdf file for the report portion of the assignments. (The CS machines have openoffice which can be
used to convert doc files to pdf if you work in Word and don’t have a Adobe PDF
printer.)
Free late days (“slip
days”): Over the
course of the term you have an allowance of four free late days for problem set
turn-ins, meaning you can accrue up to four days in late assignments with no
penalty. Late problem sets beyond this
allowance lose 50% of the total possible credit per day late. Please plan ahead so you can spend your late
days wisely. No late problem sets will
be accepted after solutions are discussed in class or posted online. Late days are timed as the assignments; we’ll
count a full additional day as having passed for submissions 1 minute to 24
hours late. If you use any late days on
an assignment, clearly include the number of days used at the top of your
write-up.
Exams:
There is an in-class midterm and a comprehensive final exam. For each exam students may use a single sheet
(8.5 x 11”) of notes.
Participation/attendance:
Regular attendance
and participation in class is expected.
If for whatever reason you are absent, it is your responsibility to find
out what you missed that day.
Reading: The reading assignments listed on the
schedule should be read before the associated class lecture.
Grading policy:
Grades will be determined roughly as follows. You can check your current grades online at
eGradebook.
·
Problem sets (55%)
·
Midterm exam (15%)
·
Final exam (20%)
·
Class
participation, including attendance (10%)
Please read the UTCS code of conduct.
Please frequently check this page for course
announcements, and use the schedule page for reading assignments, problem set
downloads, deadlines, etc.
Midterm
exam: Tuesday Oct 13 (in class) tentative
Last
class meeting: Thursday Dec 3
Final
exam: Monday Dec 14, 2:00-5:00 PM in JGB 2.218
Books
The
recommended textbook is Computer Vision: A Modern Approach, by Forsyth and
Computer vision : a modern
approach, David A. Forsyth and Jean Ponce.
Computer
vision, Linda G. Shapiro and George C. Stockman.
Introductory
techniques for 3-D computer vision, Emanuele Trucco and Alessandro Verri.
Computer vision,
Dana H. Ballard and Christopher M. Brown.
(available online)
Multiple
view geometry in computer vision, Richard Hartley and Andrew Zisserman.
Pattern classification, Richard O.
Duda, Peter E. Hart, and David G. Stork.
Machine learning, Tom M.
Mitchell.
Computer Vision:
Algorithms and Applications, book draft by Richard Szeliski
·
A
script tattle-matlab.sh
is posted on Blackboard->Course documents.
You can use it to alert you when Matlab licenses are available, in case
there’s another overload in the future.
See the script for usage instructions.
(Thanks to Jason Pepas for providing this.)
·
UTCS Computing /
Facilities web page
·
OpenCV (open
source computer vision library)
·
Weka (Java data mining software)
·
Compiled
list of image datasets
·
Object
recognition databases (list compiled by Kevin Murphy)
·
Various useful
databases and image sources (list compiled by Alyosha Efros)
·
Netlab
(matlab toolbox for data analysis techniques, written by Ian Nabney and
Christopher Bishop)
·
Oxford
Visual Geometry Group (contains links to data sets and feature extraction
software)
·
Annotated
computer vision bibliography
·
Computer vision research groups
·
Vision related links on
AAAI.org page
·
Linear
algebra review / primer by Martial Hebert