Object Recognition and Face Recognition Flashcards
(36 cards)
Define object recognition
The ability to perceive an objects physical properties and apply semantic attributes
How do humans take in information?
Register large scale shapes and patterns AS QUICKLY AS small details
Mechelli et al (2004) said what about bottom up vs top down?
That’s humans do bottom up first then top down
How do computers perceive objects?
They start with pixels (small details) “bottom up” and then build up
What are constancies?
What we see as the same, despite differences in dynamics
What’s the context effect?
Some recognition requires an appreciation of context
What did Oliva & Torralba (2007) say about object recognition?
That objects in familiar context are recognised faster
What are the 5 steps to recognise an object?
1) see basic features
2) perceive organisation of features
3) perceive shape
4) compare shape to memory for knowledge
5) is the shape familiar, tap into knowledge
What is simultangnosia?
The inability to perceive how parts fit together
List the 3 models of object recognition
1) Template Matching
2) Feature recognition
3) Structural theories
What is the template matching model?
When an individual compares the whole object to stored representations to find a match
What is the feature recognition model?
Rather than focusing on the whole shape, feature models break down into critical features
What’s Selfridges (1959) pandemonium model? Which 3 model does it come under?
It’s a metaphorical description of levels of processing patterns before the whole
Feature recognition model
What’s Marr & Nishiharas (1978) model? And what 3 models does it come under?
There are 3 levels of object recognition
1) primal sketch (edges and contours)
2) 2.5D sketch (shading and textures)
3) 3D model
The feature recognition model
What 3 things happen for recognition to occur?
1) single model axis
2) component axes
3) 3D model match (top down processing)
What is the structural theory? Using RBC theory
Biederman’s recognition by components theory (1987)
Objects consist of combinations of geons
Object recognition is viewpoint invariant
What’s greebles good for, using concavity importance?
Explores regions of concavity importance.
When presented with Greebles, individuals remained viewpoint dependent when recognising one even after training
What’s dynamics of vision?
Focus on perception of static images by static observer
What’s motion perception?
To perceive moving objects we must be able to pick up changes in stimulus over brief time intervals and recognise object in different stimuli. Humans are good at this.
What 2 things do object recognition depend on?
Sensory info and prior expectation
What plays an important role in object recognition that isn’t mentioned?
Context
What are the 4 challenges of object recognition?
Different classes of things recognised
From different angles
Partly occluded
Fully occluded
What 3 things make face recognition special?
The development
The configural processing
Mechanistic
What’s “developmental” face recognition?
Newborns are orientated towards upright face stimuli
Tendency to disappear after a month then reappear again (nakamo et al, 2014)