Face and Object Recognition

Question 1

Object recognition - how are we able to have constancy across viewpoints? = Marr’s Theory of Vision

Answer 1

Suggests we have an input image
We create a primal sketch from this
Then a 2D sketch
Then a 3D model representation of the object (which describes objects’ shapes and their relative positions independent of the observer’s viewpoint)

Question 2

Discuss Biederman’s Recognition by Components Theory

Answer 2

Object consists of a combination of Geons (almost an alphabet of shapes)
- object recognition is viewpoint invariant (the recognition of the components is driven by the properties of the visual input)

So instead of representing the kettle realistically, you would create a representation using volumetric primitives - because these geons are viewpoint invariant, it matter if that kettle is rotated or not, you’ll still recognise it

Question 3

What are the advantages and limitations of the Recognition by components theory

Answer 3

Adv
- simple
- some evidence for geons being important in object recognition
- evidence that the identification of concavities and edges is also of major importance

Limitations
- de-emphasises top-down influences
- fails to account for most within-category discriminations
- much recogniion is viewpoint dependent
- some classes do not have invariant geons but are still recognisable as members of that category (clouds)

Question 4

Multiple View recognition

Answer 4

object recognition inhernetly viewpoint dependent
to recognise an object from different viewpoints, store how it looks from many different views

Question 5

How to test these theories;

Answer 5

test participants with objects that have never seen before
- object representations are inherently viewpoint-dependent
- object recognition can be viewpoint-dependent or viewpoint invariant

FAMILIAR = viewpoint-invariant
UNFAMILIAR = viewpoint-dependent

Question 6

Holistic face processing effects

Answer 6

Face inversion effect
- inverted faces are disproportionately harder to recognise than upright faces relative to objects
Part-whole effect
- memory for a face-part more accurate when presented within the whole face
Composite effect
- obligatory influence from a task irrelevant face half on judgments about the task relevant face half

Question 7

What’s the dedicated neural substrate for facial recognition?

Answer 7

Fusiform Face Area (FFA)
- more active for faces than for other object categories
- damage to the FFA leads to prosopagnosia (inability to recognise faces)

Question 8

Contrast specialisation and domain specificity

Answer 8

Specialisation: neural substrate and/or behaviour is selective for a category of a stimuli
Domain specificity: exclusive processing of a single domain of stimuli

Question 9

FFA –> discuss

Answer 9

FFA lights up when people look at faces
FFA lights up when car experts look at cars
FFA lights up when bird experts look at birds
Therefore FFA activation is correlated with expertise –> specialisation not domain specificty

This was tested with Greebles
and this accounts for an ‘own race bias’ or ‘other race effect’ in facial recognition

Question 10

What about when expertise with faces doesn’t develop

Answer 10

People with autism tend to show abnormal face perception
- deficits in face tasks (matching or identification, emotion recognition, gaze direction judgements)
- absence of an inversion effect
- children on the spectrum have a mouth bias (eye movements) comapred to a typical eye bias
- show reduced FFA activation