Midterm 2 Flashcards Preview

PSYC 102 - sensory > Midterm 2 > Flashcards

Flashcards in Midterm 2 Deck (97)
Loading flashcards...

name of the two visual streams

ventral - temporal
dorsal - parietal
for vision and interaction in the environment


how many ganglion cell types are there and what are their names

M ganglion
P ganglion


M ganglion cells

Magnocellular layers in LGN (layers 1 and 2 - inside)
fucntion movement and low light vision
high conversion of rods goes through M ganglion cells


P ganglion cells

Parvocellular layers in LGN (3, 4, 5 & 6 - outside)
function colour, texture and depth


classic visual pathway in the brain

M ganglion - magno LGN - V1 --dorsal--- paritel lobe
P ganglion - parvo LGN - V1 ----ventral--- temporal lobe


how lesion / ablation studies work

animal trained to indicate perceptual ability
specific part of brain is ablated or removed
animal is retrained to determine which perceptual abilities remain
results reveal which portions of the brain are responsible for specific behaviours


classic ablation study for what and where pathways
-set up

animals trained on object discrimination task
-monkey shown an object
-then present with two choice task
-reward given for detecting target object
also trained on spaticial landmark discrimination problem
-monkey is trained to pick the food well next to a cylinder - so there is a spatital relationship and no object discrimination
then temporal and parietal lesions


classic ablation study

temporal lobe removed = problems in object discrimination task so what pathway
parietal lobe removed = problems in landmark discrimination task so where pathway


behaviour of patient DF and conclusions

damage to ventral pathway die to carbon monoxide posioning
could not tell orientation (perceptual orientation matching) of the slot but could actively post a letter into it (visuomotor posting)
ventral stream = what
dorsal stream = how
really prfound deficit - brain knows what it is at some level but cnanot do it


how did what where evolve

what and how


what is the inverse projection problem

an image on the retina can be caused by an infinite number of objects
fundamentally ambigious at the level of the retinal image - need to impose additional constraints
so no specific right way object recognition must work


what else makes object recognition tricky

huge variation - eg loads of pictures of dolphins, lots of inconsistencies but our brain still knows is a dolphin


what does gestalt mean

german word
configuration or pattern


according to gestalt perception...

is not built up from sensations but is a result of your brain imposing perceptual organization on incoming stimuli


gestalt principles are known as..

heuristics = best guess rules


6 gestalt organizing principles

good continuation
proximity / similarity
common fate
common region
uniform connetedness


good continuation

continuous shapes viewed as single segmented obejcts
helps us perceive a pile of rope as one continous object and not all broken up


proximity / similarity

things that are bear to each other are grouped together


common fate

things moving in the same direction are grouped together


common region

elements in the same region tend to be grouped closer together


uniform connectedness

connected a region of visual properties are peceived as a single unit



interpret images in line with top-down knowledge
eg we see faces everywhere and in everything. us imposing top down knowledge on inanimate stuff


explain gestalt laws competring

leads to an ambigous percept
shows there isn't a correct way to process images
brain just tries to apply the heuristics to disambiguate the info


charlie chaplin mask illusion

we have a lifetime experience seeing faces
never see them as concave - always projecting out
so we impose structure on the world
and so see both sides of the mask as pointing out
cannot see otherway round
top down imposition of knowledge by the brain on the bottom down world


Biederman recognition by components theory

simple computational model of object recognition
36 geons
each geon is uniquely identifiable from most viewpoints
objects can be identified if geons can be identified


evidence from Biederman

objects harder to recognise if geons are obscured, easy to recognise if can see geons


strengths of biederman

viewpoint invariance
represents 3D structure


weaknesses of biederman

complexity of representation
doesn't easily represent subtle metric differences (ie distance between the eyes)
recognition is at the level of categories (chair vs table) rather than individuals (my chair vs office chair), also trump vs someone dressed up as trump


modern day version of recognition by componenets

deep neural networks
modern machine vision
architecture for performing object recognition


how deep neural networks work

series on interconnected layers of modelled neurons
node = neuron / pop of neuron
basic processing at ealry layers eg 100000000 instead of just geons, look a lot like primary visual cortex
as progress through the layers = more and more complex representations
even modern machine learning use same general basic approach
building blocks put together to descirbe objects in detail
note this is how AI works