2. Object Recognition
• The process of identification whereby we match an
incoming stimulus with stored representations for the
purpose of identification.
• We can look at a pattern of stimulation and say “bird”
3. Levels Of Categorization Rosch (1976)
1) The superordinate level of categorization-animal is
the most general
2) The subordinate level of categorization-black capped
chickadee is the most specific
4. Levels Of Categorization Rosch (1976)
3) This midpoint between these two levels is called
The basic level of categorization- bird, horse, dog
• At the basic level- a) share similar shapes, b)requires
the same movements and postures to interact with
them, c)lead to the formation of a single mental image
• The basic level turns out to be the entry level for
recognition
5. Recognizing from the Bottom up and Top
down
• Bottom up processing-employs the information in the
stimulus itself. (we build and identify stimuli from the
bottom up.
• Top Down processing- expectations, knowledge, and/
or surrounding context to supplement the data
6. Visual Object Recognition
• Light strikes an object and we see boundaries, edges,
contours, and surfaces as well as more global
organizational processes
• These lead us to discern the shapes and objects that will
ultimately lead to recognition.
7. Effect Of Orientation And Perspective
• Objects can rotate, shift and change their orientation
8. Structural Description- Based SDB Approach
• We compare the features of the object we’ve just seen to
a description of the object’s structure stored in memory
• Often labeled as feature analysis
• -he representation stored in memory is not visually or
spatially analogous to the object being recognized
9. Structural Description- Based SDB Approach
•Viewpoint-Independent
• -recognition of the object does not depend
upon a particular view of the object; only the
component features of the object itself.
10. Recognition-By-Components (RBC)
• (Biederman and colleagues, 1991)
• The features by which we parse objects are based on
three-dimensional shapes, termed “geons”
• There are a total of 36 geons
11. Recognition-By-Components (RBC)
• These are simple shapes that combine to form more complex
shapes
• Information about the edges is extracted from the retinal
image
• Next the non-accidental features of the retinal image
(genuine features)
• We tend to parse objects in the simplest way possible
12. View-Based (VB) Approach
• A VB approach contends that objects are recognized
holistically through a process of comparison to a stored
analog, “viewpoint dependent.”
• Some theorists contend that object recognition involves both
SDB and VB mechanisms, depending on the nature of the
task.
13. View-Based (VB) Approach
• Recognition at the basic level may involve primarily SDB
mechanism, while the finer discriminations required at the
subordinate level involve primarily VB mechanisms
• From an evolutionary perspective, we have evolved a quick and
efficient recognition mechanism VB and a more attention
demanding and slow mechanism (SDB)
14. View-Based (VB) Approach
• Context has an important influence on object and scene
recognition
• Objects are more readily recognized if they are consistent
with their respective scenes, especially when the orientation is
unusual.
• Scenes are more readily recognized if they are consistent with
the objects in the scene.
16. Experiment
• Think of what you would likely see if you were looking at
a cat?
• Chances are that the visual image includes a head, pointy
ears, whiskers, four legs, a long tail, along with the
respective size and layout of these features.
17. Experiment
•How do you think a cat might feel to the
touch?
• You would likely come up with quite different
attributes such as furriness, warmth, and softness
18. EXPLORATORY PROCEDURES
• The hand movements we use for tactile identification.
• These include contour following, enclosure, lateral
motion, static contact, unsupported lifting, and pressure
19. EXPLORATORY PROCEDURES
• Contour following: moving the hands around an object to
determine shape and/or identify it
• Enclosure: Holding an object with both hands to help determine
size and shape
• Lateral motion: Gently gliding your hand back and forth over
something, often to determine texture
20. EXPLORATORY PROCEDURES
• Static Contact: Gently touching something (tapping with
fingers or palm), often used to ascertain temperature
• Unsupported Lifting: Ascertaining weight by holding
something in the palm
• Pressure: Grasping something with varying amounts of
force to determine hardness
21. The Connection Between Vision and Haptics
• Haptics aid in recognition and makes objects easier to
identify.
• For example it is theorized we have an easy time visually
recognizing an apple because over the years, we’ve
simultaneously felt them, held them, cut them into pieces,
and pulled them down from trees.
• The combination forms a visuo-haptic representation of an
apple.
22. Olfactory Recognition
• Lawson (1997)- coined the term olfactory-verbal gap which
means:
• People have a difficult time describing and identifying smells
• Research shows people are able to only correctly label 50%
of presented odors
• The difficulty in labeling odors is what may be behind the
difficulty in creating olfactory images
26. FACE RECOGNITION
• Without the ability to recognize faces we would be in a
sea of strangers
• Prosopagnosia- inability to recognize familiar faces
27. EXAMPLE
• “Farrah (1992) recounts the story of a person
suffering from this disorder who was sitting in a
country club an wondering why another gentleman
was staring at him so intently. He asked a server to
investigate only to discover the man staring at him
was his own reflection.”
29. The Thatcher Illusion
• Demonstrates that inversion has a disproportionate effect on the
recognition of faces relative to its effect on objects.
• This effect suggests that faces are encoded and subsequently
recognized holistically.
• On the other hand, objects are encoded more in terms of separate
elements
• This difference is due to the way in which the properties of a face is
configured
30. • Women have better facial recognition abilities compared
to men
• Individuals within a culture, tend to recognize same-
culture faces more readily than different-culture faces
• The right hemisphere of the brain seems to be related to
self-face recognition (contested by some)
31. First-Order Relational Information
• Diamond and Carey (1986)- in order to recognize objects, we
need information about the parts of an object and how those
parts relate to one another
WHAT FEATURES MAKE UP THE HUMAN FACE?
WHERE WOULD EACH FEATURE BE IN RELATION TO
THE OTHER FEATURES?
32. First-Order Relational Information
• For facial recognition this would involve an analysis of
the person’s facial features and the relationship among
those features.
• Simply recognizing there are two eyes above the nose
would help with recognizing it is a face not whose face it
is
33. Second-Order Relational Information
• This involves comparing the first-order analysis to facial
features of a “typical” or “average” face.
• This typical face serves as the standard to which other
faces can be compared.
• Face inversion disrupts second-order relational
information
34. HOLISTIC PROCESSING
• Faces are encoded as whole configurations and are best processed as a
complete picture
36. SPECIAL MECHANISM VIEW
• Double dissociations found between object and face recognition
• Double dissociation-when two tasks or abilities are influenced by
different variables
• This is taken as evidence that the ability to perform each task is
based on separate (brain/cognitive) processes.
• ‘Faces would be considered a special kind of object’ recognized
with (face-specific) mechanisms
37. EXPERTISE VIEW
• Superior knowledge and skill develops after extensive
practice in some domain, and there is no doubt that all
human beings could be experts in this area
38. Where is facial recognition technology used?
• Facial Recognition moving into retail
39. NETWORKS AND CONCEPTS
• Semantic networks
• A way of describing the representation of categories
and concepts
• These networks can be assessed using a category
verification task and a feature verification task
40. Category Verification Task
• how we access categorical knowledge. Participants
are asked to verify or deny simple statements like,
“A penguin is a bird” or “A robin is a bird” as
quickly as possible
41. Feature Verification Task
• How the features and categories are stored and
accessed. Participants are asked to verify sentences
like, “a cat has pointy ears” or “a cat has skin.”
43. Spreading Activation Model
• Concepts assist in understanding, predicting, and communicating
about knowledge. Example is semantic priming- the tendency for
the processing of one stimulus (e.g., yellow) to enhance or speed
up the processing of a related stimulus (e.g., lemon).
44. CONCEPTS AND CATEGORIES
• “Concepts are the building blocks of thought.” P128
• They provide us with labels that are convenient for grouping
things
• Mental shorthand that allow for quick and efficient understanding
• Concepts support new learning and are important for
communication
45. CONCEPTS AND CATEGORIES
• Natural kinds (also termed natural categories)- these are
concepts that occur naturally in the world
• Artifacts (artifact categories)-objects or conventions
designed or invented by humans to serve particular
functions
46. The classical view
• Items are classified into particular categories if they have
certain features or characteristics.
47. Problems
• Problems with the classical view-it is very difficult to
specify many categories in terms of features that are both
necessary and sufficient.
• For example, the category game, what would be common
features? Necessary and sufficient?
48. More Problems
• It cannot explain a fundamental characteristic of categorization e.g.,
graded structure are too rigid to account for the graded structure of
categories and their fuzzy boundaries. (If a member of a category has
those features then it is a member and if not then it is not a member.)
• Fuzzy boundaries-one person’s game is another person’s sport.
• For example, is bowling a sport?
• The sort of response is an example of fuzzy boundaries.
49. The Prototype Approach
• Assumes we compare objects to a best example from
the category
• Problems:
• Does to account for the graded structure and their fuzzy
boundaries
• Does not account for variability of categories in context.
50. Other Approaches
• The Exemplar Approach- proposes we think of concepts
in terms of specific examples, and accounts better for the
sensitivity of concepts to contextual factors.
• Essentialist Approaches- concepts are represented in terms
of their essence, or basic underlying nature.
51. Current Research on Concepts
• We represent concrete concepts (such as a bicycle) and abstract
concepts (e.g., jealousy) in terms of classificational
relationships
• Cross-cultural comparisons of concept representation reveal
some interesting differences in the way we think about natural
categories like “animal.”
• Categories once thought of to be amodal (e.g., free of the
senses) have been shown to include sensorimotor
characteristics