Lecture 11 - Object Recognition Flashcards

1
Q

pathways that transmit and process visual information

to produce basic features

A
Ganglion axon pathway
   • Visual thalamus – lateral geniculate 
nucleus (LGN)
   • Striate cortex (V1)
   • Hierarchical organization
   • Feature detectors

basic palette to start creating objects

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

how do we find the objects?

visual form agnosia

A

brain damage: visual form agnosia: can’t recognize objects as a certain kind of thing

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

inferior temporal lobe

A

On the basis of stroke patients with visual agnosia, it was known that this area was involved with OBJECT RECOGNITION

called the “What pathway” because it was critical for target identification tasks.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

stroke patients with parietal lobe damage

A

problems dealing with spatial relationships (being able to understand where things are in relation to other thing)

difficulty using landmarks: “‘where is the door?’ ‘it’s near the exit sign’”

can’t reference an object in the environment and then infer something from it and how they could use it, even though they could recognize it

difficulty with spatial relationships and remembering
landmarks.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Ungerleider & Mishkin (1982) tested functional separation in monkeys with ablation studies.

A
  • First, an animal was trained on some task to indicate their perceptual capacities and then measure their ability to do that task
  • Second, after training, a specific part of the brain was removed or destroyed (ablated).
  • Third, the animal was retrained and measure their ability to see how much they can get back and to determine which perceptual abilities remain.

comparison: how much of the ability could be recovered? If almost all of it could be recovered then you can infer then that area of the brain wasn’t strongly related to that function

• The results reveal which portions of the brain were responsible for specific behaviors.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Object discrimination problem

A

– Monkey was shown an object and rewarded (with food) for recognizing it as a target.

– Removal of inferior temporal lobe tissue
resulted in problems with the object discrimination task = determined the IT is “What pathway”
because it was critical for target identification tasks.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Landmark discrimination problem

A

– Monkey was trained to pick the food well
next to a cylinder.

– Removal of parietal lobe tissue resulted in
problems with the landmark discrimination task = determined this area was called the “Where pathway” as it was necessary for using landmarks and spatial relationships.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

parietal lobe

A

“where/how pathway”

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

How do we isolate visual processing streams in people?

A

go out and find large populations: what deficits are associate with which type of brain damage = See if people, as a result of brain injury or illness, tend to lose
specific functions while retaining others. [Natural experiments, because you’re not inducing anything]

looking for a dissociation

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Dissociation

loss of both?

loss of one?

A

One function is absent while another is present.

  • For example, if a person loses the ability to name objects, do they also lose the ability to determine their locations?
  • If a single injury causes the loss of both abilities, we might conclude that the abilities are related and the same brain mechanism seems associated with both abilities.
  • However, if an injury causes the loss of one only one function, perhaps different areas of the brain implement those two functions (dissociated).
  • In this case, the brain region seems associated with only one ability and is dissociated from the other.
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

dissociation example

A

you are able to identify an object but can’t do spatial reasoning with that object

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

single dissociation

example ?

A

Two functions/abilities seem to involve different
brain mechanisms.

based on one piece of damage

Alice damage to temporal lobe - can’t name objects BUT SHE CAN determine their locations in space - keeps the where info loses the what info

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

no dissociation

A

if Alice lost the temporal lobe and then couldn’t name objects or determine their locations in space

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

double dissociation

A

two functions/abilities involve different brain
mechanisms and operate independently (requires two individuals).

you need two different people to show it: show different behaviors

Bert has parietal lobe damage, can name objects find but can’t determine spatial relationships = this damage is associated with a different kind of error

Alice can do one task that Bert can’t and has different damage and vice versa

same area doesn’t seem to be related to the same functional deficits

a way of getting at which area of the brain is associated with which function

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Dissociations can also be shown in individuals with intact brains and abilities (Ganel et al., 2008).

A
  • Visual illusions provide one way of dissociating object visual judgments (perceptual identity information) from action information (‘how’ an object is used).
  • Line 2 appears longer than line 1; however, this is a perceptual illusion => converging lines give an illusion of depth => visual identification pathway is telling you based on experience, 2 has got to be bigger cause i’ve seen 100000 things like that
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Will the ‘what’ and ‘where/how’ pathways process the lines in the same way?

Ganel et al., 2008

Present the same stimuli under two tasks:

A

Length estimate task

Grasping task

same distal stimulus coxing out two different answers with the same motion

Results show subjects doing the length
estimating task falls victim to the
illusion; however, they are more
accurate during the grasping task.

support for the idea that somehow the perception, identification, and action are implemented differently in the brain

doesn’t prove independence

17
Q

Length estimate task

A

“looking at this image, tell us how long each line is, show me with your fingers how big you think it is”

measured people’s fingers

supposed to be relying on the what pathway: perception: identifying what the thing is and trying to express it in some way with actual movements

people said the 2 (farther away) looked longer

18
Q

grasping task

A

“you’re not going to reach out and grab it, but act as if you were, move thumb and forefinger”

measured fingers again

people are much more accurate = how pathway wasn’t affected by the illusion

19
Q

A patient has a stroke and afterwards has trouble using
landmarks information when driving. However, she can name
objects on a map. She likely has damage in the:

A

. Parietal lobe = Where/How Pathway

20
Q

‘Where/how pathway’

A

also called dorsal pathway (top and the back)

implicated in action or “how” we do things.

21
Q

What pathway’ also called….

A

….ventral pathway.

22
Q

Both pathways:

A

– originate in retina and continue through (specific?) ganglion cells in the LGN (separate magnocellular and parvocellular layers) then to V1 and then they send signals to temporal and parietal lobe.

– have some interconnections (V1 and beyond).

– receive feedback from higher brain areas (do eventually get top down info that limits possibilities of identifications).

23
Q

When looking for objects in the brain, we need to first
define the types of sensory codes we imagine will
represent objects.

A
  1. What are the options for the sensory codes
    (representations) ?
  2. How do those codes fit into theoretical models for object perception?

a way that the brain implements some object: takes in the distal info and says “i’m gunna keep that right here and i’m gunna keep it in some set of these neurons with some firing pattern”

24
Q

specificity coding

problem?

A

type of sensory code

One option is to have a single neuron represent each object

each face has its own neuron (Ryan Gosling Neuron)

evolutionary problems: what happens with aging, or isolated trauma that you lose your Megan Fox neuron?

25
Q

distributed coding

A

suggest that representations are instantiated in the brain by firing patterns of many groups of neurons.

This is more efficient as it allows fewer neurons to represent more stimuli. It also makes the system more resistant to damage.

sparse coding and population coding

26
Q

sparse coding

A

will use the fewest possible neurons to represent that object = more effecient

FEW GROUPS

is a version of distributed coding that begins to
place constraints on the number of neuron groups involved with a representation.

• This view suggests that fewer groups are needed for representation.

• In the extreme case (one-unit representation), you could claim that
the specificity code is a version of a sparse code.

27
Q

population coding

A

is a form of distributed coding which
suggests that a larger, more expansive set of neurons are involved in representation.

• This view argues for more groups of neurons in any representation.

28
Q

Models of object perception/representation

A

Domain specificity models

Property-based models

29
Q

Domain specificity models

which view?

expect to find?

A

Suggests that certain brain regions are dedicated to specific
categories (domains) of objects [e.g. faces, animals, tools, etc.]

  • Nativist view – more modularity, less plasticity.
  • Expect to find sparse codes.
30
Q

Property-based models

which view?

expect to find?

A

it’s not just the object, the fact that it’s a face, do we have knowledge in general of what those things mean, how we use them, and why they’re important? really complex

it isn’t just feature based, but involves a lot more about that object. how it’s meaningful, etc…

Claims that object knowledge is linked to sensory and motor
attributes of the item [note that this is more than the
physical features].

  • Constructivist/embodied cognition view – less modularity, more plasticity.
  • Expect to find population codes: an object is a complex thing: how you would use it and why you would use it
31
Q

Rolls & Tovee (1995) recorded from infero-temporal (IT) neurons in monkeys.

A

. They found that many cells responded best to faces.

An additional study found that 97% of cells in IT were selective for faces.

32
Q

facial recognition area

A

(fMRI) study

in the fusiform gyrus, directly under the IT. This is called the
fusiform face area (FFA).

Is this a face module? = kinda

33
Q

Experience dependent
plasticity in humans

Greeble experiment is important why?

A

• fMRI experiments show that training tunes FFA neurons to
respond to novel or non-face stimuli.

  • In the Greeble recognition study, FFA was found to respond to artificial creatures following 7 hours of training.
  • Neurons in FFA have also been shown to respond to cars for people who have expertise in identifying cars.
  • This is taken as evidence for the property-based theories.