CSC 487/2503, Fall 2007

CSC 487/2503, Fall 2007
Foundations of Computer Vision

General Information

Instructors	Prof. Allan Jepson
Email	`jepson at cs dot utoronto dot ca`
Office Hour	Thurs. 3-4pm, D.L. Pratt 283D
	Or by appointment.
Lectures FALL 2007	Thurs. 1-3, Lash Miller Building (LM) 157 (corner of Wilcox and St.George)
Tutorials FALL 2007	Mon. 12-1, LM 157
Course Information Sheet	Handout
Course Bulletin Board	https://csc.cdf.toronto.edu/bb/YaBB.pl
Email-Newsgroup Policy	There may be no response from 24 hours before an assignment is due, up to the actual time it is due.
Final Exam	See the exam timetable for the time, date and place.

Marking

4 Assignments	60%
Final Exam	40%	Previous years' exams

Late Policy

Assignments are due at the beginning of the tutorial or lecture on the due date (i.e. Mon. at 12:10, or Thurs. at 1:10).

The penalties for lateness are:

10% for each 24 hr. period, or any part thereof.
100% after one week.

Hard copies of late assignments should be handed into the AI Office, Pratt 283, during business hours. The date and time should be marked, and it needs to be signed for by someone in the office.

Assignments

Links to assignments and related information are provided here.

If you are new to Matlab, first read the Matlab primer and try running the code described there. Then, if you wish to set up your home machine, see the matlab/README file unpacked from the matlabVisTools.zip file (see below).

Assignment
Description Percent Posting Date Due Date Links

Asgn #1
15% Oct. 4 1:10pm, Oct. 18
at lecture. A1handout.zip
The marking is complete. Email your tutor for your mark. You can pick up your marked assignments Friday Nov. 2 in the afternoon, at the instructor's office.

Asgn #2
15% Oct. 18 1:10pm, Nov. 1 A2handout.zip
Typo(s) in handout.

Asgn #3 15% Nov. 1 1:10pm, Nov. 15
(at the beginning of the lecture) eyeFinderHandout.zip Contains starter Matlab code and test images.
If you are running Matlab version 6, use the following *.mat files:
trainSet6.mat
testSet6.mat

Asgn #4 15% Nov. 17 1:10pm, Dec. 6
(at the beginning of the lecture) The Matlab handout code pigeonTrack.zip.
The image sequence is at pigeon.zip. DO NOT DOWNLOAD the image sequence to CDF. The sequence is already on CDF in /u/jepson/pub/pigeon/.

CDF Information

By registering in this course you should have a CDF account. For information on CDF see http://www.cdf.toronto.edu/. Before going to a CDF lab, or to the course bulletin board, you need to look-up your CDF account.

Lecture and Tutorial Materials

Links to some of the lecture/tutorial notes are provided below.

Matlab tutorials are also provided to illustrate many of the concepts introduced in this course. The tutorials are installed on the CDF lab machines, and also on CS-Lab. After logging onto a CDF (or CS-Lab) machine, cd ~jepson/pub/matlab and read the README file. This file tells you how to set up matlab so that it will run the tutorials. The tutorials are located in two directories:

~jepson/pub/matlab/iseToolbox/tutorials/ (referred to as ISE below),
~jepson/pub/matlab/utvisToolbox/tutorials/ (referred to as UTVIS below).

We indicate which toolbox to look in below.

The Matlab toolboxes we use in this course are also available for download to your home computer (assuming you have Matlab at home) from the matlabVisTools.zip file (60MB). After unzipping this file, see the matlab/README file for instructions on how to set your matlab path. A second zip file containing two image sequences is needed for the motion tutorial, namely matlabVisToolsSeq.zip file (75MB). Unpack this zip file in the same directory as the previous one. It should form a new subdirectory matlab/images/seq/.

Date Topic Matlab Tutorial

Sept. 13, 07 Lecture Notes:

Introduction to Computational Vision.
Lighting and Reflectance Models
Basic Lighting and Reflection from CSC 418 notes (Ignore the OpenGL section.)
Radiometry and Reflection chapter from CSC 418 notes
Suggested Readings:

Adelson and Pentland, The perception of shading and reflectance, 1996.
Adelson, Lightness perception and lightness illusions, 2000.
See also the impossible step illusion on Adelson's illusion site.
Getting started with images in Matlab: See the introImageTutorial.m in the ISE toolbox
(See above for directions to ISE)
Extra material: Matlab M-files for Phong shading: phongDemo.m phongShade.m
The generated images: ambient light, metal spheres, shiny spheres, matte spheres
Notes on Phong shading and the peppersc.jpg image discussed there:

Sept. 20 Lecture Notes:

Image Formation. (Typo p.16)
Image Filtering and Fourier Transforms
Suggested Readings:

Notes on Fourier Analysis, Fleet and Jepson, 2005.
Camera Models Fleet and Hertzmann, 2005.
UTVIS: colourTutorial.m
ISE: linSysTutorial.m, samplingTutorial.m, imageTutorial.m
(See above for directions to UTVIS and ISE toolboxes)
Matlab demo: upSample.zip

Sept. 24 Tutorial notes on homographies Matlab demo: warpDemo.zip

Sept. 27 Lectures continue with Image Filtering notes from previous week. See the lecture for previous week.

Oct. 4 Lecture Notes:

Image Pyramids
Edge Detection
2 more pages on edge detection
Matlab Tutorials:

ISE: pyramidTutorial.m
UTVIS: cannyTutorial.m

Oct. 8 University is closed for Thanksgiving
No tutorial.

Oct. 11 Robust Estimation

See the following m-files in utvisToolbox/tutorials/lineTut/

oTensorDemo.m
houghDemo.m
robustDemo.m
robustLineScript.m
To run these, set your matlab path to include:

addpath([matlabVisRoot '/utvisToolbox/edge']); addpath([matlabVisRoot '/utvisToolbox/tutorials/lineTut']); addpath([matlabVisRoot '/utvisToolbox/tutorials/lineTut/localUtil']);

Oct. 18 Optical Flow Lecture Notes
Optical Flow Book Chapter
Lecture Notes on Mixture Models

See the following m-files in utvisToolbox/tutorials/motionTut/

introMotionTut.m
motionTutorial.m

Oct. 25 View Based Models

See the following m-files in utvisToolbox/tutorials/eigenTut/

trainEigenEyes.m
detectEigenEyes.m

Nov. 1 Local image features.
Epipolar geometry.

See the following m-files in utvisToolbox/tutorials/

SIFTtutorial/tutorial.m
3Drecon/grappleFmatrix.m

Nov. 8 Lecture notes for multiframe factorization (updated 11:00am Nov 8, see udpate for 2 updated pages).

The lecture and the Matlab tutorial are based on:

Provably-Convergent Iterative Methods for Projective Structure from Motion, S. Mahamud, M. Hebert, Y. Omori and J. Ponce, CVPR 2001.
Self-calibration and metric reconstruction in spite of varying and unknown intrinsic camera parameters, Polleyfeys, Koch and van Gool, IJCV 1999.
UTVIS: Orthographic Factorization and Projective Factorization Demos in tutorials/3Drecon/orthographic/ and tutorials/3Drecon/projective/.

Nov. 15 Tracking.

Nov. 22 Notes on Image Segmentation.

Nov. 29 Object Recognition.

Dec. 3 Tutorial Cancelled this Monday. See you next in class on Thurs.

Dec. 6 Perceivers
Motion Sequences
Scene Dynamics
Scene Dynamics Results

Assignment Description	Percent	Posting Date	Due Date	Links
Asgn #1	15%	Oct. 4	1:10pm, Oct. 18 at lecture.	A1handout.zip The marking is complete. Email your tutor for your mark. You can pick up your marked assignments Friday Nov. 2 in the afternoon, at the instructor's office.
Asgn #2	15%	Oct. 18	1:10pm, Nov. 1	A2handout.zip Typo(s) in handout.
Asgn #3	15%	Nov. 1	1:10pm, Nov. 15 (at the beginning of the lecture)	eyeFinderHandout.zip Contains starter Matlab code and test images. If you are running Matlab version 6, use the following *.mat files: trainSet6.mat testSet6.mat
Asgn #4	15%	Nov. 17	1:10pm, Dec. 6 (at the beginning of the lecture)	The Matlab handout code pigeonTrack.zip. The image sequence is at pigeon.zip. DO NOT DOWNLOAD the image sequence to CDF. The sequence is already on CDF in /u/jepson/pub/pigeon/.