Lecture 5: object recognition Flashcards
Object constancy
the ability to maintain a stable and consistant perception of an object or persion, despite changes in their physical presence, appearance or emotional state.
visual information emanting from object varies
- viewing position
- illumination
- context
What & Where pathways
Ventral stream: specialized in object perception and recognition (What we are looking for)
Dorsal stream: specialized in spatial perception (Where is the object located in space)
What & where pathways in auditory domain
ventral stream: what is the sound
Dorsal stream: where is the sound coming from
Inferior temporal cortex (IT)
neurons in the IT cortex rarely respond to simple stimuli such as lines or spots of light. they respond to more complex objects such as hands.
agnosia
inability to process sensory information even though the sense organs and memory are not defective
visual agnosia
deficit in recognising objects even when the processes for analysing basic properties such as shape, color and motion are relatively intact
Patient D.F
impairment in naming objects but when object was placed in her hand, she identified it & visual acuity was intact.
performed poorly on explicit matching task (match orientation of card and the slot) but showed normal performance in the action task (insert card in slot)
impairment in explicit matching task showed that D.F could not recognize orientation of a three dimensional object, yet when asked to insert the card in box (action task) DF performed well.
D.F impaired in the ‘what’ but not in the ‘where/how’
Lateral occipital complex (LOC)
important for object recognition
optic ataxia
patient can recognize objects, but cannot use visual information to guide their actions
lesions in the parietal cortex
example: patient cannot move towards an object to grasp it, but rather moves hands around like a person searching for light switch in the dark.
Functional near-infrared spectoscopy (FNIRS)
in infants revealed the role of LOC in processing objects with shape and texture being manipulated
LOC is sensitive to shape but not to other visual features
Binocular rivalry
Binocular rivalry is a visual phenomenon that occurs when two different images are presented simultaneously to each eye. Instead of seeing a blended image, your perception alternates between the two distinct images, meaning you can only consciously perceive one at a time.
The hierarchical coding hypothesis
elementary features are combined to create objects that can be recognized by gnostic units.
at the first level of the hierarchy depicted are edge detectors (simple cells)
these feature units combine to form corner detectors, which in turn combine to form cells that respond to even more complex stimuli such as surfaces
the grandmother cell hypothesis
also known as gnostic cell.
proposes that individual neurons or very small groups of neurons, respond specifically to highly particular objects faces or concepts.
example: there might be a single neuron that activates only when you see your grandmother’s face.
suggests:
highly specificity: certain neurons respond to specific, unique stimuli.
all or nothing response: if you lose the grandmother cell you might lose the ability to recognize your grandmother al together
The ensemble coding hypothesis
object recognition results from activation across complex feature detectors. an ensemble of cells respond to your grandmother. thus, recognition is not due to one unit but rather to collective activation of units.
predicts:
distributed representation: a network of neurons work together to represent an object or concept. each neuron in the ensemble might respond to different features of the object.
greater felxibility: losing one neuron would not erase the ability to recognize an object as recognition emerges from the combined activity of the ensemble.
Top down effects on object recognitionBar and colleagues (2006)
Used MEG to examine top-down influence on object recognition.
They compared frontal regions with those in temporal
provided evidence for top down influence of frontal regions over more posterior regions during object recognition.
Mind reading: decoding and encoding brain signals
decoding: predicting the stimulus that is being viewed based on brain activity.
encoding: how stimulus features are represented in neural activity.
Voxels (measured with fMRI) might contain neurons that are tuned to one specific orientation (aggregate of many neurons measured in that voxel).
> showed the ability to decode what images people view based on brain activity.
yet this simple model is based on V1, V2 and V3 and thus does not include information represented in higher order visual areas
Specificity of object recognition in higher visual areas
Are faces special? what was the evidence found from monkey research?
research showed some cells in the superior temporal suclus (STS) respond specifically to faces.
STS: regions that respond to faces
Functional MRI = fusiform face area (FFA): activation whilst seeing faces.
BOLD signal in FFA was stronger during intervals in which faces (vs. objects) or intact faces (vs. scrambled faces) were presented.
Electrophysiological evidence for face processing: N170
faces elicit a large negative response in the EEG signal around 170ms after stimulus onset (the N170)
cofound: we see faces much more than other objects, so could it be the level of expertise that explains the N170?
Facial perception
Chang & Tsao (2017)
they were able to determine how different each face was from the average face.
they found that neurons in the lateral (LAT) and anterior (ANT) face areas within the IT cortex showed different tuning profiles.
Are there regions that are specialized?
Parahippocampal place area (PPA) is particularly responsive when people make judgements about spatial properties or relations.
Extrastriate body area (EBA) is responsive when making judgements about the body
Testing causality:
Afraz (2006)
showed that stimulation cells in IT cortex made monkey biased towards seeing faces versus seeing a flower
Testing causality:
Moeller
Let monkey look at two faces and judge them whether they were the same or different.
When IT was stimulated the monkeys almost always judged the faces to be different.
Parivizi found that stimulating the FFA in epilepsy patients. they reported seeing distorted faces
Testing causality:
Duchaine, pitcher (2009)
applied TMS to the rOFA, rLO and rEBA whilst participants did a judgement on faces.
TMS over rOFA: participants performed worse in discriminating faces
TMS over rEBA: participants performed worse in discriminating bodies.
TMS over rLO: participants performed worse in discriminating objects.
Failures in object recognition:
visual agnosia
difficulties in recognizing objects that are presented visually or require the use of visual based representation
although patient can recognize objects, their ability to do so diminishes when the perceptual input is limited or does not include the most salient features.
damage to posterior regions in right hemisphere performed worse than either patients with anterior lesions (no agnosia) or patients with agnosia arising from the left hemisphere lesions.
