Scribd
Upload
21 views

07 - Object and Face Perception

Uploaded by

Line Li
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
21 views

07 - Object and Face Perception

Uploaded by

Line Li
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 40

Perception

Object Perception

Bas Rokers
Psychology, Center for Brain and Health
New York University Abu Dhabi
Object Perception

• View-dependent vs. view-invariant perception


• Accidental / non-accidental features
• “Natural scenes” approach
• Face perception
Introduction

What do you see?


Introduction

What do you see?


Introduction

What do you see?


Introduction

How did you recognize that all 3 images were of


houses?

How did you know that the 1st and 3rd images showed
the same house?

This is the problem of object perception and recognition,


which is solved in visual areas beyond V1.
We look with our eyes, but
we see with our brain
Visual processing can be divided into two cortical streams:

Separate pathways for


Dorsal stream
“what” and “where”
(“where” pathway)
information

Text
V1

Ventral stream (“what” pathway)


Unfortunately, we still have no idea how to solve this
problem.

Not easy to see how to make neural detectors for


houses based on retinal detectors for spots of lights
in the eye

house-detector
receptive eld?
fi
Ok for detecting a single picture of a “stick gure” house.

But this receptive eld would never work: needs to


recognize houses from different angles, sizes, colors, etc.

house-detector
receptive eld?

And how does it represent that


it’s the same house from
different directions?
fi
fi
fi
Viewpoint Dependence

View-dependent model - a model that will only recognize


particular views of an object
• template-based model
e.g.

“house” template

Problem: need a neuron (or “template”) for


every possible view of the object
- quickly run out of neurons!
- unclear how to detect two views of the
same object

One potential solution: View-independent model - a model that will


recognize an object independent of the particular viewpoint
Viewpoint invariance

• One piece of evidence for view independent models


• Perceptual system will be hesitant to adopt interpretations that
assume an accidental viewpoint.
accidental
viewpoint

Thursday, September 10, 2009


non-accidental
viewpoint

Thursday, September 10, 2009


Gestalt Principles

• Gestalt: In German, “form” or “whole”


• Gestalt psychology: “The whole is greater than the sum of
its parts.”

• Opposed to other schools of thought (e.g., structuralism) that


emphasize the basic elements of perception

structuralists:
• perception is built up from “atoms” of
sensation (color, orientation)
• challenged by cases where perception
seems to go beyond the information
available (eg, illusory contours)
Gestalt Principles

Gestalt grouping rules:

a set of rules that describe when elements in an image


will appear to group together
Gestalt Principles

Good continuation: A Gestalt grouping rule stating


that two elements will tend to group together if they
lie on the same contour
Gestalt Principles

Good continuation: A Gestalt grouping rule stating that


two elements will tend to group together if they lie on the
same contour
Critique of Gestalt Principles

What’s wrong with the Gestalt principles?


• heuristics (rules of thumb)
• not implementable as algorithms
• not clear how useful their predictions are

Alternative: Natural Scenes Approach

1. Empirical studies: go measure the statistical properties


of real images of natural scenes
2. Generative Models: make an explicit mathematical
model of natural scene properties
Natural Scenes: empirical approach Geisler et al 2009

• Take a bunch of photos of natural scenes


• Count up statistics on contours, gure/ground, etc
fi
Natural Scenes: empirical approach Geisler et al 2009

Measuring the true


probability
distribution for
“good continuation”

Measure of the
probability that two
edges belong to the
same object, as a
function of edge
orientation and position
Natural Scenes: empirical approach Geisler et al 2009

Could do the same thing (i.e., go measure them) for other


properties of natural scenes: color, parallelism, symmetry,
surroundedness, etc.

Better approach than just introspecting about what we see!


Summary

• View-dependent vs. view-invariant perception


• Accidental / non-accidental features
• Gestalt principles (heuristics)
• “Natural scenes” approach
Perception
Face Perception

Bas Rokers
Psychology, Center for Brain and Health
New York University Abu Dhabi
Outline

• Specialized vision: face and place


perception

• The ‘what’ pathway and grandmother cells

• Neural areas involved in face/place


perception

• Perceptual disorders: prosopagnosia and


object agnosia
Visual processing can be divided into two cortical streams:

Separate pathways for


Dorsal stream
“what” and “where”
(“where” pathway)
information

V1

Ventral stream (“what” pathway)


Object Recognition

Object recognition is fast. (on the order of 100-200 ms)


Suggests operation of a (mostly) feed-forward process.

Feed-forward process: carries out a computation one


neural step after another, without need for feedback
from a later stage to an earlier stage

(Still hotly debated, but it’s agreed that there’s not much time for
feedback to affect processing).
Inferotemporal (IT) cortex show high selectivity to people /
things, independent of viewpoint
• e.g., “Jennifer Anniston neuron”
• Single-cell recording from human brains!

Quiroga et al 2005
“Grandmother cell” - idea of a single neuron responsible for
representing some complex object (eg, granny)
• idea of the grandmother cell predates these experiments
• considered “idea that could never work” (how could you have a different
neuron for every possible object you know how to recognize?)

Could a single neuron be responsible for recognizing your


grandmother?

Bad idea: what if you kill


off that neuron during a
long drinking session?
Figure 8.17 fMRI reveals that different pieces of the cortex are activated by faces and by places
There are areas in the brain specialized for face
perception

• Fusiform face area


 Area in the fusiform gyrus that responds preferentially
to faces

• Parahippocampal place area


 Area in temporal lobe that respond preferentially to
images of places (such as houses)
Object & Face Recognition
• Face recognition seems to be special and different from object
recognition
• Agnosia: failure to recognize objects despite being able to see them
• Prosopagnosia: inability to recognize faces
Object & Face Recognition

• Double dissociation: When one perceptual function can


be damaged without affecting the other
• Chuck Close
• Artist with
prosopagnosia
Evidence that face perception is a ‘visual expertise’ task

• We can induce prosopagnosia-like symptoms in normal


observers

• Face recognition depends only in small part on


recognizing individual features

• Depends in large part on the recognition of


relationships between features

• These relationships can be disrupted by what seem


inconsequential modifications of visual stimuli
Face Recognition
Face Recognition: not view-invariant
Summary

• Specialized vision: face and place


perception

• The ‘what’ pathway and grandmother cells

• Neural areas involved in face/place


perception

• Perceptual disorders: prosopagnosia and


object agnosia

You might also like