Scribd
Upload
58 views

Lecture 4

Vision as inverse graphics - Build an internal "model" of the world. How do you recognize the banana? How can you get a computer to do this? List a set of steps the computer has to do at a high level but make sure they are correct.

Uploaded by

satyabasha
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
58 views

Lecture 4

Vision as inverse graphics - Build an internal "model" of the world. How do you recognize the banana? How can you get a computer to do this? List a set of steps the computer has to do at a high level but make sure they are correct.

Uploaded by

satyabasha
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Michael J.

Black CS143 Intro to Computer Vision


Introduction to Computer Vision
Michael J. Black
Sept 2009
Lecture 4: Introduction
Michael J. Black
Review
Formalizing the tricks of artists.
Vision as inverse graphics
Build an internal model of the world
Too many parameters to do this in practice
We have to make some assumptions
Represent our beliefs about what is typical
(priors); e.g. light from above
Maybe our model isnt as good as we think
CS143 Intro to Computer Vision
Michael J. Black
Goals
What happens when the cues are weak?
What can be extracted bottom up?
Begin talking about object detection and
recognition
CS143 Intro to Computer Vision
Michael J. Black CS143 Intro to Computer Vision
Computer Vision
Is it more than
inverse graphics?
How do you recognize
the banana?
How can you get a
computer to do
this?
Michael J. Black CS143 Intro to Computer Vision
Object Recognition
Barrow and Tenenbaum
Michael J. Black CS143 Intro to Computer Vision
Subjective Contours
Kaniza Triangle.
Perceptual
Organization
Michael J. Black CS143 Intro to Computer Vision
I. Rock, The Logic of Perception, 1983.
The types of
junctions give
cues about
surfaces,
occlusion, and
light.
Michael J. Black CS143 Intro to Computer Vision
What goes with What
So maybe we need some sort of perceptual
organization process that tells us what low-level
measurements might group together.
Then what? How do can we recognize objects?
Michael J. Black CS143 Intro to Computer Vision
Assignment 0 In Class
Formalize the problem of detecting the the known objects in this
scene; that is their 3D pose (translation and rotation).
List a set of steps the computer has to do at a high level but
make sure that these steps are implementable.

You might also like