Computer Vision, CSE 559

1
Computer Vision, CSE 559

• Me Robert Pless
• Office hours, TBD
• Also profpless YahooIM, AIM.
• Website http//www.cs.wustl.edu/pless/559
• No textbook. Assigned readings and outside
  information will be on the website.

2
What is computer vision?

• Trucco and Verri computing properties of the 3D
  world from one or more digital images
• Stockman and Shapiro To make useful decisions
  about real physical objects and scenes based on
  sensed images
• Ballard and Brown The construction of explicit,
  meaningful description of physical objects from
  images
• Forsyth and Ponce Extracting descriptions of the
  world from pictures or sequences of pictures
• Sometimes called the pixels to predicates
  problem, a special case of the AI signals to
  symbols problem.

3
What isnt Computer Vision

• Image processing, pixels to pixels.
• Pattern Recognition, symbols to symbols.
• Photogrammetry, pixels to kilometers

4
Graphics vs. Vision, andWhy is (Computer) Vision
Hard?
Comparing to very, very good biological
implementation.
5
Why is (Computer) Vision Hard?
6
Seeing motions thats not there
7
Seeing things that arent there
8
Slide from David Kriegman, UCSD
9
Class Structure

• Class will be arranged as 2-3 week units,
  covering 5 different topics.
• 1 or 2 papers as assigned reading
• Homework assignment
• Programming project (2-3 person groups)

10
Topic 1 - FACES

• Face detection
• What is a face?
• What is a detection?

11
(No Transcript)
12
Topic 2 3D world reconstruction

• What measurements to make on images.
• What can you reconstruct from 2 images?
• Linear Algebra,
• 3x3 matrices/inverses
• Geometry
• Projections
• Robust Statistics
• RANSAC
• PhotoTourism Movie (Highlights)

13
Topic 3 Tracking

• Introduction to video.
• How to keep track of an object as it moves.
• What is an object?
• How can it move (vary, change)

14
(No Transcript)
15

• Find unusual motions.

16
Topic 4 Recognizing Actions In Video

• What is an action

17
(No Transcript)
18
Special topics

• Image segmentation
• Medical image analysis
• Content Based Image Retrieval

19
History and Carpe Diem

• Vision has a checkered history.
• But a combination of factors makes it possible to
  do things today that really havent been done
  before.
• Some things only work when you have (and can
  index and use) millions of images.

20
(No Transcript)
21
Expectations

• This is a graduate level computer science class.
• I expect you to be able to program, debug, etc.
• I will never look at your code, therefore, you
  can write your programs in whatever language you
  want Java, C, Matlab, Cobol, Lisp, Machine
  Language, perl, gawk (?), Pascal, Basic, ProLog
  (!), Scheme, VHDL, Javascript, Flash.
• Some of the above languages are easier than
  others.
• I expect you to follow the unive

22
Grading

• Project/homework portfolio.
• About every two weeks, there will be an
  assignment that will have a problem set part and
  a project part.
• 40 Projects
• will mostly be done in small groups, and turned
  in on the web.
• 30 In class midterm and final exam
• 30 Final project, portfolio evaluation.

23
Topic 1 - FACES

• Face detection
• Scale, rotation and pose.
• Do we need to solve for these to know there is a
  face?

24
(No Transcript)
25
Faces
26
Assignment.

• Hello world, for computer vision.
• Write a program to take any image (really!, any
  image) and return the location of the two eyes.
• Reading over next two weeks (available at)
• http//vision.ai.uiuc.edu/mhyang/face-detection-su
  rvey.html
• PAMI paper, Beginning of paper through section
  2.2 Section 3.3 (at end).
• Viola Jones paper.