HUL 211 Perception of real world objects - The neuroscience by sharathiiith


Object Perception and Memory Lecture Series

More Info
									Perception of real world objects

          Snehlata Jaswal

Object perception

            Lateral Occipital Complex
Object perception begins in V1 which extracts simple features that are
common to all images, e.g., lines. It ends in areas that extract very complex
features particular to a few related objects, e.g., faces.

In V2 & V3, the upper and lower visual fields are separated by V1.

In the object perception area (the Lateral Occipital Complex, LOC), object
parts in the upper and lower visual fields are brought together again but not
as yet those in the left and right visual.

Which features are those of the object involves combining cues such as
color (from the double opponent cells in V1 blobs) & form (from simple &
complex cells in the V1 spokes).

In the lateral occipital complex, elements of objects are extracted from the
background using both color and form and bound together. LOC codes that
something is an object part. Lesions of LOC result in visual agnosia, the
inability to perceive all objects through vision.
                         HUL 211 OBJECT PERCEPTION AND MEMORY
        The inferior temporal cortex
While LOC codes that something is
an object part, areas of the inferior
temporal (IT) cortex code a whole
particular object (e.g. a face).

In the inferior temporal cortex cells
respond selectively to a particular
class of object, e.g. faces, hands,
animals etc.

Cells that respond to the shape of
hands do not respond
to faces.

Cells are tuned to particular
instances of object, e.g. a particular
                          HUL 211 OBJECT PERCEPTION AND MEMORY
      The inferior temporal cortex

IT has a columnar

Cells within a column are
activated by the same object.

Neighbouring columns
respond best to objects of a
similar shape as in a and b.

                    HUL 211 OBJECT PERCEPTION AND MEMORY
       The inferior temporal cortex
In the inferior temporal cortex
    the response of cells is the
    same independent of

   i) the location of the
      object’s image on the
      retina because cells
      have large bilateral
      receptive fields.

   ii) the size of the image

   iii) the cue the defines the
        objects shape (e.g.
        lines, color, texture,

                       HUL 211 OBJECT PERCEPTION AND MEMORY
       Difference between V1 and IT
Some images look somewhat                     Other images look very different
similar but represent different               but are the same thing. These fire
things. These fire many of the                very different cells in V1 but the
same cells in V1 but different cells          same cells in IT.
in IT.

                          HUL 211 OBJECT PERCEPTION AND MEMORY
Two hemispheres – two ventral pathways?
The ventral pathway is often described and modeled as a single entity. It is
actually split between two hemispheres and very little is known about how they
cooperate to give rise to integrated visual perception.

Data from monkeys whose corpus callosum is cut, shows that each ventral
pathway can analyze half of the visual scene on its own, in large part
independently of what is going on in the other one. Further, each hemisphere has
its own limited pool of computational resources. Human studies have also
suggest that during bilateral stimulation each hemisphere analyzes mainly the
contra-lateral hemi-field.

In split-brain patients with trans-section of the corpus callosum, bilateral search
arrays can be inspected twice as fast than by control subjects. In normal subjects,
the interference between the two hemispheres might occur in the visual
system itself or in other structures.

Information from each ventral pathway might then be combined in structures such
as the prefrontal cortex. However, the mechanisms leading to the perception of a
unified object are not yet fully understood.

                          HUL 211 OBJECT PERCEPTION AND MEMORY
              The two hemispheres
The two sides of the brain
  are not equal.

The left side usually
  specializes in language,
  i.e., the recognition of
  words and sentences.
  The right side usually
  specializes in objects
  with spatially organized
  features, e.g., faces.

Perhaps this explains why it
  is sometimes difficult to
  associate a name with a
                      HUL 211 OBJECT PERCEPTION AND MEMORY
Perceiving real world objects is very fast
Brain processes involved in object recognition are remarkably fast. You can
recognize a most objects in less than 100ms.

Thorpe, Fize, and Marlot (1996) showed that decisions regarding presence or
absence of animals in a natural visual scene flashed for 20 ms could be observed as
ERP negativity in the frontal areas of human participants around 150 ms after
stimulus onset.

VanRullen and Thorpe (2001) demonstrated that whereas categorization of targets
(animals vs. vehicles) in visual scenes flashed for 20 ms could be observed in ERPs
75-80 ms after stimulus onset, the difference between targets and non-targets could
be observed only after 150 ms, reaching significance at 160 ms.

Roelfsema, Tolboom, and Khayat (2007) used multi-unit recordings from the brains
of monkeys and observed that features are registered in 48 ms, bound as figures
distinct from the background at 57 ms, and relevant figures are selected over
irrelevant distracters after 137 ms.

Whereas encoding of the stimulus dimensions may take less than 50 ms and the
differentiation of the object from the background happens shortly thereafter, the
categorization of objects as targets or irrelevant takes place only around 150 ms, to
be followed by further refinement by attention.
Grill-Spector and Kanwisher (2005)
                For each exposure duration, accuracy was
                lower and reaction time longer on a within-
                category identification task (e.g.,
                distinguishing pigeons from other birds) than
                on a perceptual categorization task (e.g.,
                birds vs. cars).

                Strikingly, at each exposure duration,
                subjects performed just as quickly and
                accurately on the categorization task as they
                did on a task requiring only object detection:
                By the time subjects knew an image
                contained an object at all, they already knew
                its category. Object detection and
                categorization is almost simultaneous.

                Although substantially more processing is
                required to precisely identify an object than
                to determine its general category, it takes no
                longer to determine an object’s category
                than to simply detect its presence.
            Canonical representations
Seminal research by Palmer, Rosch, and Chase (1981) examined how object
viewpoint information was accessed in a number of different tasks, and found
evidence for consistently preferred viewpoints. For example, during goodness
judgments of photographs of objects over different viewpoints, three quarter
perspectives (in which the front, side, and top surfaces were visually present)
were usually ranked highest.1 The “best” view was also the perspective imagined
when given the name of the object, the view most photographed, and enabled
fastest naming of objects. The consistencies across observers and across tasks
led Palmer, Rosch, and Chase (1981) to term this view the “canonical

Konkle and Oliva (2010) examined how internal object representations represent
visual size information. They found that real-world objects have a consistent
visual size at which they are drawn, imagined, and preferentially viewed.
Importantly, this visual size is proportional to the logarithm of the assumed size of
the object in the world, and is best characterized by the ratio of the object and the
frame of space around it. They term this consistent visual size information the
canonical visual size. Both short and long term memory for objects is also biased
towards this normative canonical size.

                             HUL 211 OBJECT PERCEPTION AND MEMORY
      Object representations in LTM
Konkle, Brady, Alvarez, and Oliva (2010) explored long term memory for objects.
Subjects viewed 2,800 object images with a different number of exemplars
presented from each category, studying each image for 3 secs, interleaved with
800 ms fixation cross. At test, they indicated which of 2 exemplars they had
previously studied.

Memory performance was high and remained quite high (82% accuracy) with 16
exemplars from a category in memory, demonstrating a large memory capacity
for object exemplars.

However, memory performance decreased as more exemplars were held in
memory, implying systematic categorical interference.

Object categories with conceptually distinctive exemplars showed less
interference in memory as the number of exemplars increased.

Interference in memory was not predicted by the perceptual distinctiveness of
exemplars from an object category, though these perceptual measures predicted
visual search rates for an object target among exemplars. These data show that
observers’ capacity to remember visual information in long-term memory depends
more on conceptual structure than perceptual distinctiveness.
                          HUL 211 OBJECT PERCEPTION AND MEMORY
         Thank you


To top