CSE252A
Computer Vision I

Tuesday, Thursday 11:00 am - 12:20 pm
Warren Lecture Hall (WLH), Room 2204

 

Class Mailing List: cse252a-fa10
Class Web Board: QuickTopic

Class web page: http://cseweb.ucsd.edu/classes/fa10/cse252a

 

News:

12/02: The final exam is on Wed (12/08) from 11:30-2:29 in WLH 2204 (sample questions). Good luck!

11/30: There was a mistake in assignment 4 -part D which has been fixed.

11/23: Assignment 4 is posted.

11/10: Assignment 3 due date is extended to 11/23. There will be no class this Thursday (11/11) due to the Veterans Day.

11/07: Assignment 3 is posted.

10/26: The credits for problem 1 in assignment 2 are increased (see below for details).

10/22: Assignment 2 is posted.

10/18: The content of Assignment 1 was mistakenly changed in the past few days which includes a few more questions.

            Please verify that you are working on the right version (link below) which only have 3 questions in Problem 3 and 4 questions in Problem 5.

10/13: Next week, the TA will have an additional office hour on Mon (10/18) 4:00 - 5:00 pm (in the same room).

10/12: Link to a supplemental material for Assignment 1 is updated.

10/08: Assignment 1 is posted.

09/30: Prof. Kriegman has a plan to be out of town on 09/30. The lecture will be presented by the TA.

09/24: Office hours for Prof. Kriegman and the TA have been set. See below for details.

09/23: The first day of class.

 

Course Information:

Instructor: David Kriegman
Office: EBU3B, Room 4120
Phone: (858) 822-2424
Email: kriegman at cs.ucsd.edu
Office Hours: Tue 1:30 - 2:30 pm

 

TA: Cuong Tran

Email: cutran at cs.ucsd.edu

Office: EBU3b, Room B240A (CSE Building Basement)

Office Hours: Tue 3:00 - 4:50 pm

Class Description:
Comprehensive introduction to computer vision providing broad coverage including low level vision (image formation, photometry, color, image feature detection), inferring 3D properties from images (shape-from-shading, stereo vision, motion interpretation) and object recognition. A companion course, CSE252B, Computer Vision II is taught in the Winter quarter. 4 units.

Required Text:
"Computer vision: A Modern Approach," David A. Forsyth, Jean Ponce, Prentice Hall, ISBN: 0130851981

Supplemental Text:

“Computer Vision: Algorithms and Applications”, Richard Szeliski, A draft of the book, which is currently being written, is available at: http://szeliski.org/Book/.

Prerequisites:
Linear algebra and Multivariable calculus (e.g., Math 20A & 20F), programming, data structure/algorithms (e.g., CSE100). Probability can also be useful.

Programming:
Assignments will include both written problem sets and programming assignments in Matlab. Students can either purchase the Matlab student edition or use copies available on University machines such as are available in the APE Lab.

Grading:
Assignments: 60%
Final Exam: 40%

Late Policy:
Written homework will be due in class and accepted thereafter with a penalty of 10% per day starting from the due date. Programming assignments will have a hand-in procedure described with the assignment, and also has a 10% per day late penalty. No assignments will be accepted after the graded assignments have been returned or the solutions have been released.

 

Assignments

Homework 0 [Due 10/05]: Getting Started with Matlab

• Matlab Tutorial Code Snippets
• trees.tif photo.jpg even.jpg odd.jpg

 

Homework 1 [Due 10/21]: Camera & Lenses, Rigid Transformations, and Homography

• house_model.m
• img1.jpg  img2.jpg  img3.jpg  img4.jpg  img5.jpg  img6.jpg
• Supplemental material: More details of the homography estimation process can be found here
• There was a typo in Equation (1) which has been corrected. It should be x' = Mx not x = Mx

 

Homework 2 [Due 11/04]: Radiometry and Photometric Stereo

• Data for the programming assignment part: Synthetic Images, Specular Sphere Images, Pear Images, Squash Images
• The credits for problem 1 are increased (from 2 to 4pts) due to the amount of computation in comparison with other problems. The goal of these paper and pencil problems is to derive the equation for calculating irradiance. In case the resulting integral is hard to solve analytically (e.g. problem 1.b), just simplify it as much as you can then solve it numerically (e.g. using the quad function in Matlab).

 

Homework 3 [Due 11/18  11/23]: Epipolar Geometry and Dense Stereo Matching

• Input Images. Either save the links below or download all the images at once, all_ims.zip.

  Left	Right	True Disparity
  			DOTS
  			BARN1
  			TSUKUBA

• Supplemental reading : "A Maximum Likelihood Stereo Algorithm", by Cox, Hingorani, Rao, and Maggs,

 

Homework 4 [Due 12/03]: Optical Flow and Video Stabilization

• There was a mistake in part D regarding warping images for stabilization. Please refer to the updated version of the assignment.

• Input Images:  hw4_images.zip

• Input videos for stabilization: clip1.avi, clip2.avi

• There is no class on the due date so just put the hard copy part of your submission in Cuong's mailbox (CSE Building, 2nd floor) by the end of the due date.

   

  Synth	Sphere	Corridor

 

• Supplemental reading : Trucco & Verri pp178-211.

 

 

Syllabus:

 “

[ Note that this Syllabus is tentative & subject to change.  Readings denoted F&P are from Computer vision: A Modern Approach “and those denoted by RZ are from Computer Vision: Algorithms and Applications.” ]

Week	Date / Link to lecture notes	Topic / Readings
1	Sep. 23 Linear algebra review Random variables review	Intro to Computer Vision
2	Sep. 28	Human Visual System, F&P sec. 1.3. RZ 1-18
	Sep. 30	Rigid Transformatoins SE(3), SO(3), Camera & Lenses, F&P Sec. 2.1, F&P Chap. 1, RZ. 29-45
3	Oct. 5	Perspective, Affine, orthographic projection, F&P 2.2, 2.3, RZ. 46-59
	Oct. 7 Partb	Radiometry (Irradiance, Radiance, BRDF), F&P Chapter 4; Supplemental reading Glassner Chapter 13 (requires campus network connection)
4	Oct. 12	Special BRDF's, Light Sources, Photometric Stereo, F&P, 5.2-5.4;  RZ, 60-72, Supplemental reading Glassner pp 726-752
	Oct. 14 Irradiance example problem	Photometric stereo RZ 580-583
5	Oct. 19	Illumination Cones, Belhumeur, Kriegman, What Is the Set of Images of an Object under All Possible Illumination Conditions?, IJCV 28(3), 1998, 245-260
	Oct. 21 white background version)	Color; F&P, Chap. 6, RZ 8-89, Some on-line sources include: Basics of Color, A FAQ on Color
6	Oct. 26	Color, Dichromatic model, RZ 67
	Oct. 28	Filtering F&P Chap. 7, 8, RZ.  101-1.22
7	Nov. 2	Edges RZ 238=249
	Nov. 4	Epipolar Constraint and Stereo I, F&P Sec. 10.1, RZ 530-544
8	Nov. 9	Stereo II, Dynamic Programming, Chapter 11, 545-548, 552-556
	Nov. 11	Optical Flow, Trucco and Verri, pp. 178-194, RZ 4381-414
9	Nov. 16	Infinitesimal structure from Motion, Trucco and Verri pp. 195-202, 208-211
	Nov. 18	Tracking, F&P Chap.17, RZ 235-237, 282-284, 551-552, 605-609
10	Nov. 23	Statistical Pattern Recognition, F&P 22.1-22.3
11	Nov. 30	Support Vector Machines & Kernel Methods, F&P Sec. 22.5, 22.8
	Dec. 2	Appearance-based Recognition and  Model-based recognition, F&P Chap. 18, RZ 655-722

Notes and Links:

Programming languages:
The primary language will be Matlab. . Click here for Serge Belongie's Matlab resource links.

Other excellent textbooks:
Introductory Techniques for 3-D Computer Vision, Trucco and Verri (textbook for CSE152)
An Invitation to 3D Vision: From Images to Geometric Models, Ma, Soatto, Kosecka and Sastry, Springer Verlag, 2003, ISBN 0-387-00893-4 (textbook for CSE252B)

Some useful links:
Camera Calibration Toolbox for Matlab (Bouguet)
Microsoft Camera Calibration Code (Zhang)
Intel OpenCV
CVonline
The Computer Vision Home Page
Handy Math reference: MathWorld

Related classes at UCSD: CSE 190-A, CSE 166, CSE 252B