Sebastian Thrun CS223B Computer Vision, Winter 2005 1 Stanford CS223B Computer Vision, Winter 2005 Lecture 1 Intro and Image Formation Sebastian Thrun,

1Sebastian Thrun CS223B Computer Vision, Winter 2005

Stanford CS223B Computer Vision, Winter 2005

Lecture 1 Intro and Image Formation

Sebastian Thrun, Stanford

Rick Szeliski, Microsoft

Hendrik Dahlkamp, Stanford

2Sebastian Thrun CS223B Computer Vision, Winter 2005

Today’s Goals

• Learn about CS223b

• Get Excited about Computer Vision

• Learn about Image Formation (tbc)

3Sebastian Thrun CS223B Computer Vision, Winter 2005

Administrativa

• Time and LocationTue/Thu 1:15-2:35, Gates B03SCPD Televised (Live on Channel E5)

• Web sitehttp://cs223b.cs.stanford.edu

Class Email list (announcements only)[email protected]

• Class newsgroup (discussion)su.class.cs223b (server: news.stanford.edu)

4Sebastian Thrun CS223B Computer Vision, Winter 2005

People Involved

• You! (63 students)

• Me!

• Rick Szeliski, Microsoft

• Hendrik Dahlkamp:

5Sebastian Thrun CS223B Computer Vision, Winter 2005

6Sebastian Thrun CS223B Computer Vision, Winter 2005

The Text

7Sebastian Thrun CS223B Computer Vision, Winter 2005

Course Overview

• Basics– Image Formation and Camera Calibration– Image Features

• 3D Reconstruction– Stereo– Image Mosaics

• Motion– Optical Flow– Structure From Motion– Tracking

• Object detection and recognition– Grouping– Detection– Segmentaiton– Classification

8Sebastian Thrun CS223B Computer Vision, Winter 2005

Course Outline

• http://cs223b.stanford.edu/schedule.html

9Sebastian Thrun CS223B Computer Vision, Winter 2005

Goals

• To familiarize you with basic the techniques and jargon in the field

• To enable you to solve computer vision problems

• To let you experience (and appreciate!) the difficulties of real-world computer vision

• To get you excited!

10Sebastian Thrun CS223B Computer Vision, Winter 2005

Requirements• Attend + participate in all classes except at

most two• Turn in all assignments (even if for zero

credit)• Pass the midterm exam • Successfully carry out research project

– Jan 31: selection– Feb 14: Interim report– March 8/10: Class presentation– March 15: Final report

• No exceptions!

11Sebastian Thrun CS223B Computer Vision, Winter 2005

Grading Criteria

• 10% Participation

• 30% Assignments

• 30% Midterm exam

• 30% Project

(35% of all students received an A in CS223b-04)

12Sebastian Thrun CS223B Computer Vision, Winter 2005

Today’s Goals

• Learn about CS223b

• Get Excited about Computer Vision

• Learn about image formation (tbc)

13Sebastian Thrun CS223B Computer Vision, Winter 2005

Computer Graphics

Image

Output

ModelSyntheticCamera

(slides courtesy of Michael Cohen)

14Sebastian Thrun CS223B Computer Vision, Winter 2005

Real Scene

Computer Vision

Real Cameras

Model

Output

(slides courtesy of Michael Cohen)

15Sebastian Thrun CS223B Computer Vision, Winter 2005

Combined

Model Real Scene

Real Cameras

Image

Output

SyntheticCamera

(slides courtesy of Michael Cohen)

16Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 1:Stereo

See http://schwehr.org/photoRealVR/example.html

17Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 2: Structure From Motion

http://medic.rad.jhmi.edu/pbazin/perso/Research/SfMvideo.html

18Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 3: 3D Modeling

http://www.photogrammetry.ethz.ch/research/cause/3dreconstruction3.html

19Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 4: Classification

http://elib.cs.berkeley.edu/photos/classify/

20Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 4: Classification

http://elib.cs.berkeley.edu/photos/classify/

21Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 5: Detection and Tracking

http://www.seeingmachines.com/facelab.htm

22Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 6: Optical Flow

David Stavens, Andrew Lookingbill, David Lieb, CS223b Winter 2004

23Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 7: Learning

Andrew Lookingbill, David Lieb, CS223b Winter 2004

Demo: Dirt Road

24Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 8: Human Vision

25Sebastian Thrun CS223B Computer Vision, Winter 2005

Example 8: Human Vision

26Sebastian Thrun CS223B Computer Vision, Winter 2005

Excited Yet?

27Sebastian Thrun CS223B Computer Vision, Winter 2005

Computer Vision [Trucco&Verri’98]

28Sebastian Thrun CS223B Computer Vision, Winter 2005

Today’s Goals

• Learn about CS223b

• Get Excited about Computer Vision

• Learn about image formation (tbc)

29Sebastian Thrun CS223B Computer Vision, Winter 2005

Topics

• Pinhole Camera

• Orthographic Projection

• Perspective Camera Model

• Weak-Perspective Camera Model

30Sebastian Thrun CS223B Computer Vision, Winter 2005

Pinhole Camera

*many slides in this lecture from Marc Pollefeys comp256, Lect 2

-- Brunelleschi, XVth Century

31Sebastian Thrun CS223B Computer Vision, Winter 2005

Perspective Projection

A “similar triangle’s” approach to vision. Notes 1.1

Marc Pollefeys

32Sebastian Thrun CS223B Computer Vision, Winter 2005

Perspective Projection

x

fZ Z

fXx

XO

-x

33Sebastian Thrun CS223B Computer Vision, Winter 2005

Consequences: Parallel lines meet

• There exist vanishing points

Marc Pollefeys

34Sebastian Thrun CS223B Computer Vision, Winter 2005

Vanishing points

VPL VPRH

VP1VP2

VP3

Different directions correspond to different vanishing points

Marc Pollefeys

35Sebastian Thrun CS223B Computer Vision, Winter 2005

The Effect of Perspective

36Sebastian Thrun CS223B Computer Vision, Winter 2005

Implications For Perception*

* A Cartoon Epistemology: http://cns-alumni.bu.edu/~slehar/cartoonepist/cartoonepist.html

Same size things get smaller, we hardly notice…

Parallel lines meet at a point…

37Sebastian Thrun CS223B Computer Vision, Winter 2005

Perspective Projection

fZ Z

fXx

XO

-x

38Sebastian Thrun CS223B Computer Vision, Winter 2005

Weak Perspective Projection

f

Z

O-x

ZZ

XconstZ

fXx

Z

39Sebastian Thrun CS223B Computer Vision, Winter 2005

Generalization of Orthographic Projection

yY

xX When the camera is at a(roughly constant) distancefrom the scene, take m=1.

Marc Pollefeys

40Sebastian Thrun CS223B Computer Vision, Winter 2005

Pictorial Comparison

Weak perspective Perspective

Marc Pollefeys

41Sebastian Thrun CS223B Computer Vision, Winter 2005

camera theoflength focal

depth

scoordinate world,,

scoordinate image,

f

Z

ZYX

yx

Summary: Perspective Laws

1. Perspective

2. Weak perspective

3. OrthographicYconstyXconstx

Z

fYy

Z

fXx

YyXx

42Sebastian Thrun CS223B Computer Vision, Winter 2005

Limits for pinhole cameras