Object Recognition Flashcards
what is object recognition?
processes involved in identifying objects in the visual field
what is mental chronometry?
the time course of a cognitive process
Donder’s simple reaction time is …
is the fastest time possible to make a response after seeing the stimulus
Donder’s choice reaction time is …
the time to push one of two buttons in response to a stimulus
The difference between both of Donder’s reaction times shows what?
it reveals how long it takes to make a response
What are the two key attentional systems involved in object recognition?
the ventral system
the dorsal system
What is Marr’s computational theory
it describes different representations that are constructed in object recognition
1) grey level
2) primal sketch
3) 2.5 sketch
4) 3-D model representations
In Marr’s computational theory what is grey level?
it is the first representation of the stimulus constructed
it compares the intensity of light across the retina, in order to identify edges, bars and terminations (known as place tokens)
In Marr’s computational theory what is raw primal sketch?
it is the second representation of the stimulus which is reconstructed
it groups place tokens into patterns. This grouping is based on spatial pattern, size4 and density of the reoccurring images
In Marr’s computational theory what is 2.5 sketch?
it is the third representation of the stimulus reconstruction
it is a viewer centred construction, vectors have been attached to the edges to indicate orientation
In Marr’s computational theory what is the 3-D model
it is the fourth representation of the reconstructed stimulus
it allows the the viewer to imagine objects in a rotated position, the observer can recognise the object from different perspectives
Summary of Marr’s approach…
object recognition is a computational process
it is a bottom up approach, that does not consider top-down and bottom up processing working in parallel
Biederman’s recognition by components theory (RBC)
Objects are comprised of basic shapes known as geons
there are 36 geons and these can be combined in different ways
allowing us to recognise objects
The steps of Biederman’s object recognition ..
1) edge extraction provides a line drawing of the object
2) segmenting a visual object into parts
3) deciding which edge information can be invariant depending on angle
4) geons constructed from these invariant properties
5) compare neon construction with object representations stored in LTM
Summary of Biederman’s (geon) components theory
it emphasises bottom-up processing
object recognition is viewer invariant (objects can be recognised from all angles)
evidence for biedermans component theory
Biederman (1987)
participants were presented with degraded lone drawings
if concavities (corners/junctions) were missing objects were harder to recognise
Evidence for Bierderman’s Geons …
asked observers to decide whether objects presented in quick succession had the same name
when a neon was changed the task performance worsened
Biederman’s problem of viewpoint invariant recognition …
viewpoint invariance occurs with familiar objects, however, viewpoint dependence is needed for novel objects
Evidence that novelty is viewer dependent …
participants spent 7 hours learning to identify grebes
two grebes were then presented in quick succession and the observer had to decide whether they were the same or not despite the orientation changing
they found that the time taken to identify grebes increased as the orientation difference also increased
new theories think object recognition is both invariant and dependent, but when are these viewpoints used?
invariant = categorisation
dependent = identification
Does complexity influence viewer perspective?
Vanrie et al (2002) argued that complexity influenced whether object recognition was invariant or dependent
what area of the brain is involved in object recognition?
inferotemporal cortex there are some neurons that have high invariance and others have low invariance
different regions of the infernotemporal cortex is specialised for different categories (e.g faces/places)
what did neuroscience find out about the processing of object recognition?
found evidence for ventral bottom up processing but also that the dorsal stream (top down processes) are involved.
*** visual system has forward and backward projecting neurons
evidence for top down processing …
ambiguous figures having two interpretations
when you prime a particular bias more likely to see that bias in the figure
neuroscience evidence for top down processing …
when you mask drawings, the orbitofrontal cortex is activated first before the recognition pathway is activated
