chapter 3 questions

explain why detecting edges is fundamentally important for object recognition

our ability to recognize objects depends on our high spatial resolution central vision

• edges define the boundaries of objects

describe figure/ground perception

separating the figure (the main object in the image) from the ground (the foreground/texture of the image)

what are bistable images

figures that can be perceived in 2 ways and your perception changes back and forth between the 2 perceptions

• necker cube

compare and contrast algorithms and heuristics

algorithms

• methodical procedures that are slow (many steps)

• always get you the right answer

• used by computers

• not constructive

heuristics

• best guesses that are faster

• more error-prone

• used by humans

• constructuve

why do computers sometimes have difficulty recognizing objects

• computer can’t always identify incomplete edges due to contrast

• computers can’t recognize shadows in scene segmentation

describe the pandemonium model of object recognition

a hierarchical model where there are feature demons providing input to cognitive demons which then feed a decision demon which decides what is present by looking at which cognitive demon is most active

• each demon is like a neuron in the brain

what are the major problems with the pandemonium model of object recognition

• the whole thing breaks down when we try to find something like the decision demon

• there aren’t any neurons or areas in the brain that are solely responsible for deciding what we see

discuss the template theory of object recognition

a theory that we have a template or complex pattern detector that matches up with a particular letter

what is the major problem with template theory of object recognition

• even when we’re just talking about letters, there are too many variations to have a template for each one

• we would need a template for every object which isn’t possible

discuss the recognition by components theory of object recognition

states that objects can be made up of simple, 3D objects called geons

discuss the feature integration model of object recognition

a stage theory

• pre-attentive stage

  • brain unconsciously evaluates all objects in parallel

  • immediately processes basic textural information

• focused attention stage

  • must consciously search for things in a serial manner

describe the computational approach to object recognition

stage theory

• primal sketch

  • basic difference in light between objects and their backgrounds are computed

• 2 ½ D sketch

  • based on primal sketch

  • constructs a representation of the world that is viewpoint dependent

discuss the “structuralist” and “Gestalt” approaches to object recognition

structuralist

• states that what we perceive is merely the sum of its individual parts

gestalt

• states that the perceptual whole is greater than the sum of its individual parts

• states that there are many principles by which the brain organizes perceptual information

  • based on the brains experience with the visual world

what are the Gestalt grouping principles

• the law of good figure/simplicity

  • we perceive the simplest possible pattern

• law of similarity

  • things that are alike are grouped together

• law of good continuation

  • lines are seen following smoothest possible path

• law of proximity

  • objects are closer to each other are grouped together

• law of pragnanz

  • our brain constructs the simplest, most statistically likely perception of all possible alternatives

what is the law of pragnanz and why is it important

states that our brains construct the simplest and most statistically likely perception of all possible alternatives

what happens when we trick the brain into applying the law of pragnanz inappropriately

visual illusions

• we can trick our brain into seeing things that aren’t really there based on misapplication of law of pragnanz

how does the visual system deal with the aperture problem

• parallel processing

  • brain looks at receptive fields all at the same time

• convergence

  • when multiple neurons send info to a single neuron

    • leads to larger receptive fields and more complex tuning

how does the visual system deal with the occlusion problem

the brain has to construct a perception of what should be behind the object

• relatability

  • if two lines are relatable, the brain can easily connect them

• non-accidental figures

  • arrow junctions, y-junctions, t-junctions in corners that make up boundaries of objects

why does the perception of faces seem to be special

• faces provide a lot of info about someone

• helps us identify people

• no other area of the brain that is able to process and recognize faces

discuss the area of the brain involved with facial perception and recognition

fusiform face area

• in right inferior temporal lobe

what happens if the fusiform face area is destroyed

• balint’s syndrome

  • people can’t use visual info to guide their movements

• visual agnosia

  • the inability to identify objects using visual info

compare and contrast the what and where/how pathway

what pathway

• ventral stream → temporal lobe

• involved with object identification

• damage = visual agnosia

where/how pathway

• dorsal stream → parietal lobe

• involved with object location & visually guided movements

• damage on both sides = balint’s syndrome

discuss evidence that the brain area involved with facial recognition may be more generally involved with expert object recognition

greeble study

• trained people to become experts at recognizing greebles

• measured activity of the fusiform face area using fMRI

• found that as people become better at recognizing greebles, more activity would be observed in the FFA

• FFA can be trained to help us become experts at recognizing other types of objects