object recognition

Question 1

Inverse projection problem:

Answer 1

The same object can project different images on the retina

Question 2

Viewpoint invariance:

Answer 2

Our ability to recognize an object from any viewpoint

Question 3

middle vision:

Answer 3

Is a loosely defined stage of visual processing between basic feature extraction and object recognition and scene understanding

Identification of edges and surfaces

Grouping of different regions of an image into object

Question 4

Illusory contours:

Answer 4

We can perceive contours even though nothing changes from one side to the other … so can cats

Question 5

Gestalt grouping rules :

Answer 5

Describe when elements in an image will appear to group together

Question 6

Good continuation is:

Answer 6

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

Question 7

Good continuation can be detected by neurons with:

Answer 7

aligned receptive fields

Question 8

Middle vision rules:

Answer 8

common properties allow us to group parts of an image together

Question 9

Gestalt rules:

Answer 9

good continuation, similarity, proximity, parallelism, symmetry, meaningfulness, and familiarity are examples of Gestalt grouping rules

Question 10

dynamic grouping properties (2):

Answer 10

o elements that share a common fate (move together) tend to group together
o elements that are synchronised tend to group together

Question 11

How do we use the features that we have extracted from a scene to apply the rules?

Answer 11

one metaphor for how our brain does this is that it’s like committees coming to a consensus decision
different and sometimes competing principles are involved and our perception reflects the consensus that emerges

Question 12

committee rules

Answer 12

respect physics and avoid accidents

Question 13

an ambiguous figure is a visual stimulus that permits two or more possible interpretations of its identity or structure. In the case of ambiguity, our perceptual committees

Answer 13

Tend to obey the laws of physics

Question 14

an accidental viewpoint is a position that produces some regularity in the visual image not present in the world
our committees assume that viewpoints :

Answer 14

are not accidental

Question 15

The Bayesian approach allows us;

Answer 15

To calculate the probability of a particular hypothesis (interpretation) given an observation (stimulus)
P(H|O)

Question 16

Bayesian techniques provide:

Answer 16

A formal way to model the perceptual decision making process

Question 17

ground =

Answer 17

background

Question 18

figure =

Answer 18

object of interest

Question 19

How do we decide what parts of an image belongs to the (back)ground and what parts belong to the figure (object of interest)?

Answer 19

gestalt principles:

size: smaller regions are likely to be seen as figure

symmetry symmetrical regions are likely to be seen as figure

parralelism: regions with parralel contours are likely to be seen as figure

meaningfulness: if we recognize a shape it is more likely to be seen as figure

extremal edges: if edges are shaded such that they seem to recede into the distance, they tend to be seen as figure

relative motion (depth): if one region moves in front of another, the closer region is figure

surroundedness: a surrounding region (border) is likely to be ground

Question 20

goals of middle vision

Answer 20

o Bring together that which should be brought together
o Split asunder that which should be split asunder
o Use what you know
o Avoid accidents
o Seek consensus and avoid ambiguity

Question 21

figure-ground assignment can be explained by

Answer 21

Gestalt Principles

Question 22

• meaningful shapes are likely to be recognised as

Answer 22

“figure”

Question 23

Naïve template theory:

Answer 23

The proposal that the visual system recognizes objects by matching the neural representation of the image with a stored representation of the same “shape” in the brain

Question 24

Structural description:

Answer 24

A description of an object in terms of the nature of its constituent parts and the relationships between those parts

Question 25

the problem with receptive fields:

Answer 25

the problem with templates is that we would need a different template for every size, orientation, and style of the same “thing”
that’s a lot of templates

Question 26

recognition by components

Answer 26

-Biederman (1987) proposed that we recognise objects by the identities and relationships of their component parts

-their component parts are constructed from a finite set of geometric icons, or geons

-we can create a wide variety of objects from geons

-and this theory would allow us to perceive objects with viewpoint invariance – ie the description of the object isn’t affected by the angle we are looking from

Question 27

problems with structural descriptions (recognition by components)

Answer 27

these descriptions can be too broad – how to we distinguish between individual cups or suitcases?
Geons aren’t always the best primitives to describe things
our ability to recognise objects isn’t completely viewpoint invariant
the farther an object is rotated away from a learned view, the longer it takes to recognise

Question 28

A possible solution to the problems with structural descriptions

Answer 28

perhaps we have different object recognition processes that depend on the category level

entry-level category : the label that first comes to mind when we see an object

subordinate-level category : a more specific term for the object

superordinate-level category: a more general term for the object

Question 29

__components can perfectly explain visual object recognition

Answer 29

• neither simple template matching nor recognition by components can perfectly explain visual object recognition

Question 30

*we take longer to recognise objects

Answer 30

that are rotated away from familiar viewpoints

Question 31

Occlusions (i.e., objects hiding other objects):

Answer 31

are common in the environment.

Question 32

By area V4, cells are interested in stimuli such as

Answer 32

fans, spirals, and pinwheels

Question 33

FFA: Fusiform Face Area. Responds
to

Answer 33

faces more than other objects

Question 34

PPA: Parahippocampal Place Area.
Responds preferentially to

Answer 34

places, such as pictures of houses

Question 35

EBA: Extrastriate Body Area.
Specifically involved in

Answer 35

the perception of body parts

Question 36

MT: Middle Temporal area.
Specialized for:

Answer 36

motion processing

Question 37

Figure–ground assignment:

Answer 37

The process of determining that some regions of an image belong to a foreground object (figure) and other regions are part of the background (ground)

Question 38

Reliability:

Answer 38

The degree to which two line segments appear to be part of the same contour

Question 39

Global superiority effect:

Answer 39

The properties of the whole object take precedence over the properties of parts of the object

Question 40

Nonaccidental feature:

Answer 40

A feature of an object that is not dependent on the exact (or accidental) viewing position of the observer

Question 41

T junctions:

Answer 41

indicate occlusion. Top of T
is in front and stem of T is in back

Question 42

Y junctions:

Answer 42

Indicate corners facing the
observer

Question 43

Arrow junctions:

Answer 43

: Indicate corners facing
away from the observer

Question 44

Prosopagnosia:

Answer 44

An inability to recognize faces

Question 45

Agnosia

Answer 45

A failure to recognize objects despite being able to see them

Question 46

Double dissociation:

Answer 46

When one perceptual function can be
damaged without affecting the other