networks and faces Flashcards
deep neural network (DNN)
‘machine learning’ in ai; computer programmed to learn something (e.g. object recognition)
levels of object recognition
entry, subordinate (specific), superordinate (general)
4 challenges for DNN in replicating human object recognition
- fooled by image manipulations that humans aren’t
- can’t recognize objects in multiple ways (levels of object recognition)
- subordinate categories may come first for unusual category members
- entry level shifts down for experts
structural description (geons) is for which category of object recognition?
entry level
viewpoint representation is important for which category of object representation?
subordinate level
example for an unusual category member that might be recognized at subordinate level first?
ostrich
example for when entry level shifts down for experts?
recognizing specific bird rather than just ‘bird’
face inversion effect
recognition is more difficult for inverted faces; distortions go unnoticed
holistic processing
analysis of entire object/scene; upright faces but not inverted faces or other objects
composite face effect
when faces mixed up: slower, less accurate id; not occur for misaligned or inverted faces
visual agnosia
cannot interpret visual sensory information; impairment in perception/recognition w/ relatively intact low-level vision
prosopagnosia (definition & name 2 types)
selective inability to perceive/recognize faces; usually associative; apperceptive
associative prosopagnosia is what and involves what part(s) of brain?
selective inability to recognize faces; R or bilateral anterior temporal lobe
apperceptive prosopagnosia is what and involves what part(s) of brain?
selective inability to perceive/recognize faces; R or bilateral occipitotemporal/fusiform gyrus
what prosopagnosia did the man who mistook his wife for a hat have?
apperceptive prosopagnosia
T/F: prosopagnosia can have a developmental/congenital form with normal brain structure
true
T/F: all people with prosopagnosia can recognize facial expressions
false; sometimes they can recognize facial expressions
fusiform face area (FFA) (function and location)
(may process) face identity selectively; bottom surface of brain
posterior superior temporal sulcus (fSTS)
(may process) face expression selectively; lateral near occipital lobe
occipital face area (OFA)
(may process) face identity and expression; bottom surface of brain in between FFA and fSTS
two sides of face processing controversy?
- faces require special processing
- faces processed by object recognition mechanisms
support for 1. faces require special processing?
face inversion effect, composite face effect, prosopagnosia, fMRI selective activation of FFA
support for 2. faces processed by object recognition mechanisms
-non-face inversion effects (experts of handwriting)
-non-face holistic processing
-prosopagnosia specific to subordinate categories for faces and other objects
-fMRI activation of FFA by greebles
-FFA does subordinate level discrimination fine-tuned by experience
where pathway location
superior occipital and parietal cortex
what pathway location
inferior occipital and temporal cortex
lesion to __ result in difference with visual object discrimination
what pathway
lesion to __ result in difference with visual object position
where pathway
patient DF had no activation of __
lateral occipital lobe (object id task)
patient DF had __ and could not interpret visual sensory info
visual agnosia
what was patient DF evidence for?
what and how streams; double dissociation between grasping and perception
the opposite of what patient DF had?
optic ataxia
optic ataxia
inability to reach/grasp object even if visual processing is intact
evidence for what and how streams? (visual-geometric illusions)
-ebbinghaus (dot surrounded by more dots)
-horizontal-vertical illusion (T)
-tilt illusion
-müller-lyer illusion (arrows)
T/F: perceptual judgements show less distortions in awareness and visually-guided actions show a larger illusion
false; visual system is affected more than action system
