GASB05H3 Chapter Notes - Chapter 109-119: Parahippocampal Gyrus, Extrastriate Body Area, Prosopagnosia
Parts and Wholes:
-Global superiority effect: the finding in various experiments that the properties of the whole
object take precedence over the properties of the parts of the object
othe first goal is to carve the retinal image into large-scale objects
- Don Hoffman and Whitman Richards notes that when one blob is pushed into another, a pair of
concavities is created in the silhouette of the resulting two-part object.
oAccording to aspects of physics, for this
reason boundaries should be marked by
valleys (concavity), rather than bumps
We are inclined to describe the object
into 2 parts, not into 3 parts
Summarizing Middle Vision:
- Early-vision: processes gave us the local features in the visual world
- Middle-vision processes began the work of understanding what those local features might be
telling us about the state of the world. The work of the middle-vision processes discussed so far
in this chapter can be summarized in 5 principles:
oBring together that which should be brought together
Gestalt group principles
Similarity
Proximity
Parallelism
Symmetry
Processes that complete contours and objects when they are partially hidden
behind occluders
Relatability heuristic
oSplit asunder that which should be split asunder
Complementing the grouping principles are the edge-finding processes that
divide regions from each other
Figure-ground mechanisms separate objects from the background.
Texture segmentation processes divide one region from the next based on
image statistics
oUse what you know
2-D edge configurations are taken to indicate 3-D corners or occlusion borders,
and objects are divided into parts based on an implicit knowledge of the physics
of image formation
oAvoid accidents
Avoid interpretations that require the assumptions of highly specific, accidental
combinations of features or accidental viewpoints
oSeek consensus and avoid ambiguity
Every image is ambiguous
There are always multiple, even infinite physical situations that could generate a
given image
Using the first 4 principles, the committees of middle vision must eliminate all
but one of the possibilities, thereby resolving the ambiguity and delivering a
single solution to the perceptual problem at hand
find more resources at oneclass.com
find more resources at oneclass.com
From Metaphor to Formal Model
- Knowledge about regularities in the world can constrain the
interpretation of ambiguous sensory input can be modelled in
mathematical ways
oBayesian approach: way of formalizing the idea that our
perception is a combination of the current stimulus and our
knowledge about the conditions of the world
What is and is not likely to occur
Bayes’ theorem: P(A/O) =P(A) x P (O/A)/ P (O)
Lets us calculate the probability that the world is in a
state (A) given a particular observation (O)
oWhen an observation is made, multiple hypothesises are derived.
oHow does our visual system decide that one of these hypothesis is the best?
oThe Bayesian approach says to look for 2 things:
Prior probability: Before you look at anything, how likely is what you are
proposing?
How consistent is each hypothesis with observation?
oImage:
3 hypothesises:
A: we are looking at two green rectangles
B: suggests that this is a green square with a C on it
C: proposes that this is a green square with a B on it
oIdeas of this sort can be reduced to formal, mathematical equations with the Bayesian
approach
The Motherfucking Article in the Text: Material Perception
- Material perception is lit:
oMaterial perception IS NOT about recognizing what the objects/things are made out of
Well fuck me
oOne can use the visual properties of a material to perform 2 RELATED but not IDENTICAL
tasks
can categorize the material and can estimate how the “thing” will behave
will the cloth be itchy on the skin? Or soft?
This makes no sense to me wtf man shit I hate this course
- How to perform tasks of material perception?
oVisual cues are created by the way light interacts with physical materials
oEasier to go from physics to the creation of a visual stimulus, but NOT from visual
stimulus back to physics
Ex. Plastic
There are many kinds of plastic like plastic bags, plastic containers
Computer graphics may be doing a good job translating plastic bags to
plastic containers BUT there is no simple rule that will let you categorize
those things as plastic
How are you doing this tricky tasks?
oOne may be learning that some features correlate to a certain
substance generative model
find more resources at oneclass.com
find more resources at oneclass.com
oWhen one creates a generative model, they estimate how the
object/substance would look in different conditions (dimmer
light, etc.)
oThis helps them create an internal definition for that substance
oOne can identify many types of materials if given exposure for 40 milliseconds
The material can be real or fake
Object Recognition:
- The what pathway: from V1 into the temporal lobe
oIn V1, cells respond best to lines and edges in very specific areas of the visual field
oIn V2, we get early steps from local feature to objects
oCells have sensitivity to border ownership and illusory contours
oBy V4, cells appear to be interested in complex attributes
No set of stimuli
determined
oVariation in the orientation,
color, or other basic features in
a systematic way can occur
Variation in object
shape is harder to
imagine
In the image to the
right, the darkness of the circles indicates how well a specific cell responded to
each of these items
The cells seem to have a taste for stimuli with a feature pointing to the
right
Connection to object perception is shown in the
image to the right
A cell in V4 that responded well to sharp features
pointing to the right would respond well to (a), but
not (b)
Since that sharp point is the accidental product of occlusion
This V4 cell might respond to a stimulus like the one in the image right above,
but it will NOT respond well to a stimulus in the first set of images I posted in
this section
This is because that black object does not have a feature pointing right
oAccident of occlusion
The border ownership rules not become part of the process of understanding
object shape
- Shit gets complex when we get farther into the temporal lobe
- The image of the car:
oNeurons might be sensitive to shapes that can be put together into a recognizable
object
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Global superiority effect: the finding in various experiments that the properties of the whole object take precedence over the properties of the parts of the object: the first goal is to carve the retinal image into large-scale objects. We are inclined to describe the object into 2 parts, not into 3 parts. Early-vision: processes gave us the local features in the visual world. Middle-vision processes began the work of understanding what those local features might be telling us about the state of the world. The work of the middle-vision processes discussed so far in this chapter can be summarized in 5 principles: bring together that which should be brought together. Processes that complete contours and objects when they are partially hidden behind occluders. Relatability heuristic: split asunder that which should be split asunder. Complementing the grouping principles are the edge-finding processes that divide regions from each other.
