Object Recognition (1) Flashcards
object agnosia
can see the object but cannot put the whole thing together
does object recognition = object perception
no
apperceptive agnosia
perceptual representation
associative agnosia
identification
meaning
prosopagnosia
face blindness
can tell it is a face but cant tell who it belongs to
what happens when the pattern recognition system is damaged
agnosia
pattern recognition and form perception theories
template matching
feature analysis
structural approach
template theory
- requires a huge number of templates to account for all variation
- works well in constrained environments
- calculates correlation between template and input
can template theory be complete solution?
no
features theory
- objects are composed of separable features
- tested with feature search’s
- pop-out search
physiological evidence for features
neuron recording
cognitive visual search evidence for features
conjunction search (multiple feature search) search asymmetries
are people faster or slower if they have to look for a conjunction of features
slower
reaction time independent of display size for
single feature search
search asymmetry
- Easy to find a tilted line against vertical ones, harder to find a vertical line against tilted ones
- Easy to find a vertical line against tilted ones
- Tells us what the features are
- Easier to find the “tilt” than the “not tilt” indication that a feature we use is “tilt” not “absence of tilt” (gap not absence of gap)
structural theory
Our mental models of objects are made up of components called geons
is it easier to identify objects when geons are intact or distrupted
intact
geon distruption task
- Every other edge or line is deleted to create complementary images (images contain different geons)
You see complete geons essentially so can identify object faster
are some things hard to distinguish based on geons?
yes
apple vs. orange
are faces special?
probably
hollistic not feature processing
we are highly specialize
support for faces being special
inverted faces
composite face effect
opposite of prosopagnosia
super recognizers
overall how do we recognize objects
features
templates
geons
word recognition
example of object recognition
how to study reading
reduce stimulus presentation effect
what do they use to test reading
tachistoscope
how to measure reading
recognition threshold (ms)
percentage recognized
word frequency
repetition priming
- viewing a word once greatly helps you read it the next time
- priming is like a warm up for your brain
- has a much bigger affect on low frequency words
context effects
context helps us to recognize objects, and change what we perceive
context effects bottom up processing
its not all putting features together
context effects top down processing
context, knowledge about the world matters too
word superiority effect
it is easier to perceive a letter when it is part of a word than when it is presented on its own
is reading just putting individual letters together
no people are actually more accurate when the letter was presented in a whole world, than by itself
reading error
tend to misread less common sequences as more common one (over-regularization)
feature net
each detector has its own baseline activation level (resting level)
word frequency
higher frequency words are recognized faster (higher baseline levels)
repetition priming
- The current activation level of a word node is still active above baseline when the corresponding stimulus word is repeated hence it takes relatively less additional activation energy and less time for this word node to reach threshold the second times
- Baseline itself may be affected only after LOTS of encounters
well-formedness
faster to read well formed strings
recover from errors in sensory and perceptual processing
The same mechanisms that lead to misperceptions, also allows us to recover from errors in sensory and perceptual processing
robust pattern recognition
mechanisms help to explain the word superiority effect as well
proofreading
writing primes you read slow read backwards take a break from paper input misperceived
