Object perception - HLP1 Flashcards
What are Gestalt’s grouping principals? (5)
- similarity
- proximity
- closure
- good continuation
- common fate
What is figure-ground?
An area bound by closure is seen as a separate object. Contours are seen as belonging to one object at a time
What is the purpose of Marr’s model of recognition?
proposing reasons for why Gestalt’s principals happen
What are the 3 stages of Marr’s model of recognition? Briefly describe what happens in them
- primal sketch (2D representation of luminance)
- 2 1/2 D sketch (depth, orientation, shading, texture, motion binocular disparity - viewpoint dependent)
- 3D model (viewpoint invariant)
What is the principal axis?
The biggest cylinder that makes up an object
How do you recognise an object according to Marr? (3 stages)
- Identify the principal axis
- work through smaller and smaller cylinders
- match with 3D models in memory to the closest match
What 2 things does Marr’s model of recognition predict?
- visibility of the principal axis is important
- different orientations are equally easy to recognise
What is a problem with Marr’s model of recognition?
Many objects are hard to recognise if they are upside down or the principal axis is pointed towards the viewer
What are the 4 stages in Biederman’s recognition by components?
- edge extraction (surface characteristics)
- detect arrangement of edges (don’t alter with view)
- segment object into components (parts)
- determine GEON type for each component
What are GEONS?
Different shapes that are found in objects and their arrangement is used for object recognition
What are 3 problems with Biederman’s model?
- doesn’t differentiate between objects in a class (e.g. 2 guinea pigs)
- doesn’t use surface pattern
- says recognition is viewpoint invariant but evidence says otherwise
What brain regions are in the object processing pathways?
occipital cortex (v1) passes down to the anterior pole of the ventral temporal cortex, or to more dorsal regions to the posterior parietal cortex
What region is necessary for the object discrimination task in monkeys?
Temporal lobe - ventral pathway (what)
What region is necessary for the landmark discrimination task in monkeys?
parietal lobe - dorsal pathway (where)
What damage causes object agnosia?
Ventral lateral temporal cortex
What are the symptoms of object agnosia? How do they do in the object posting task?
inability to recognise objects (but can describe/draw from memory well)
Can’t match the hand well but can post the object well
What happens in Titchener circles illusion?
people think the middle circles are different sizes but they are actually the same - influenced by the surrounding array
but they will reach with the correct grip
What can you conclude from Titchener circles illusion?
Perception is affected by the illusion but grip aperture is not, so object analysis for action is separate from conscious perception of the object
What are the symptoms of optic ataxia? (4)
- difficulty with visually guided reaching tasks
- difficulty reaching in the right direction
- difficulty positioning fingers correctly towards an object
- little relationship between grip aperture and object size
Where in the brain is affected in optic ataxia?
parietal cortex - dorsal visual processing
What is the role of the Lateral Occipital Cortex (LOC) in object recognition?
Identity representation
Location-tolerant object information and object-tolerant location information
what did Hubel and Wiesel find in the visual cortex of the cat and monkey?
- simple cells that orient to a bar of light at a certain angle
- complex cells that respond to an oriented bar of light anywhere
- hypercomplex cells that respond to bars of light of the correct length
What do each of these respond to: V1, V2, V4, posterior inferotemporal cortex, anterior inferotemporal cortex?
- v1 = edges
- v2 = contours
- v4 = colour and shape
- PIT = simple features
- AIT = specific elaborate features
How does cell selectivity position itself in the inferotemporal cortex?
- columns that will like a specific thing, e.g. a star column
- posterior cells are orientation and size specific
- anterior cells are less sensitive to orientation and size
What are the 3 layers in the hierarchical model of object recognition?
- V1 = simple stuff
- V2 = more complex
- PIT = most complex
What are 5 positives of the hierarchical model of object recognition?
- anatomically and physiologically plausible
- based on earlier hierarchical models
- copes with viewpoint dependence/independence
- incorporates theories of learning - familiarity = recognise from novel viewpoints
- copes with multiple objects and in different contexts
How would bottom-up processing work in object recognition?
start from low level features and move up to more detail until you reach the correct shape to match something in memory
How would top-down processing work in object recognition?
start with the general context and find things that would fit there and match the best one
How can top-down and bottom-up processing work together to have optimal object recognition? What models suggest this?
context can inhibit or excite certain things to make it easier for the bottom-up process to solve any ambiguity
bidirectional processing models
Which is more important: top-down or bottom-up processing in object recognition? Why?
Top-down as there are more descending connections than ascending
What is the word superiority effect?
detecting a letter is easier when it is in a word
At what age do babies start to show object permanence?
6 months
Where in the brain activates when perceiving the similarity of objects?
ventral visual cortex
What order is the hierarchy of how we group objects as similar in Cichy’s study?
- colour
- shape
- background, free-arrangement
What is a convolutional neural network?
AI pattern recognition for images, based on biological processes in the visual cortex of animals
uses probabilities that something matches something else
