Object & Face Recognition Flashcards
What is object recognition?
Object recognition is central to human perception in identifying:
- How useful something is.
- How something can be utilised.
- Whether something is harmful or not.
What are challenges of object recognition?
- Binding problem.
- Incomplete or camouflaging objects.
- Categorisation vs. identification.
- Contextual and bottom-up influences.
- Object invariance.
What is the binding problem?
Sensory information from different sensory channels is brought together and unified, to form a coherent stream of perceptual experience.
What is perceptual constancy?
The retinal images for familiar objects change under different visual conditions.
- The brain must be able to maintain some constancy in its perception of the outside world, and it does.
What is Marr’s Theory of Object Recognition (1982)?
Primal sketch
- 2D description of edges, intensity, and contours.
- Identifies boundaries and regions (similar to Gestalt’s figure-ground).
- Viewpoint-dependent.
2 ½-D sketch
- Incorporates shading, texture, motion, and binocular disparity for depth and orientation.
- Viewpoint-dependent (spatial locations from the viewer’s position).
- Not yet fully 3D.
3D sketch
- Viewpoint-independent 3D object/scene representation.
- Matches object shape with stored descriptions.
- Top-down processing confirms/denies shape.
- Results in symbolic visual representation.
What are the limitations of Marr’s theory?
- He focused excessively on bottom-up processes.
- He assumed that “vision tells the truth about what is out there”; in fact, there are numerous exceptions.
- Many processes proposed are incredibly complex computationally.
- He de-emphasised the major role expectations and knowledge play in object recognition.
What is the Recognition-By-Components Theory by Biederman (1987)?
- Objects are recognised by “geons” (geometric ions), the basic units of objects.
- Geons are easily distinguishable, view-invariant, and recognisable even if parts are missing.
- Objects are broken into geons and their spatial relationships, then stored in memory.
- Geons are based on non-accidental (invariant) properties.
- Recognising geons allows us to identify objects.
What are the limitations of Biederman’s theory?
- The theory focuses primarily on processes triggered by the stimulus input.
- It accounts only for fairly unsubtle perceptual discriminations.
- It’s difficult to extract geons from real images.
What is pareidolia?
We are primed to see faces in every corner of the visual world.
Do newborn babies prefer to view faces?
Turati et al. (2002)
1-4 day-old infants look longer at displays with more elements in the top half, regardless of whether they are face-like or not.
What are the three types of processing that have been distinguished in face recognition?
- Detecting the first-order relations that define faces.
- Processing second-order relations (spacing among features).
- Configural/holistic processing or relational processing (putting features together into gestalt).
What is Bruce & Young’s Model (1986)?
It separates processing for familiar vs unfamiliar faces.
- Separate routes for Expression Analysis and Facial Speech Analysis.
Two routes:
- Identity route
- Expression/characteristic route
- Route 1 (Familiar Faces): Face Recognition Unit (FRU) matches faces to memory, activating Person Identity Nodes (PINs) for identity info.
- Route 2 (Unfamiliar Faces): Analyzes expressions and lip movements.
- Face Classification: Identifies traits and activates FRU if familiar.
What evidence supports Bruce & Young’s Model?
- Young et al. (1986): People recognise familiar faces faster than identifying politicians or naming faces, supporting sequential processing (FRU → PIN → NG).
- Young et al. (1993): 20% of brain-damaged patients could recognise faces but not expressions, showing separate processing routes.
- Campbell et al. (1986): Two stroke patients showed double dissociation:
Patient A could recognize facial expressions but not lip read.
Patient B could lip read but not recognize facial expressions.
What is the evidence of the existence of holistic processing?
- Part-whole effect
- Distortion effect
- Composite effect
What is the inversion effect?
A phenomenon where identifying upside-down faces compared to upright faces is much more difficult than doing the same for non-facial objects.
What is the expertise hypothesis?
- Holistic or configural processing is not unique to faces but should be found for any objects of expertise.
- The Fusiform Face Area should be highly activated when observers recognise the members of any category for which they possess expertise.
What is the Fusiform Face Area (FFA)?
A brain area which exhibits a greater response to faces than other objects.
What does the Extrastriate Body Area (EBA) do?
Responds preferentially to images of human bodies and body parts, compared to objects.
What does the Parahippocampal Place Area (PPA) do?
It appears to process visual information related to places in the local environment.
What is the Lateral Occipital Cortex (LOC)?
A brain area that is especially responsive to shapes.
What is the Visual Word Form Area (VWFA)?
An area selective for the recognition of written strings relative to other categories.
What is visual agnosia?
A condition where a person can see but cannot recognise or interpret visual information.
What is apperceptive agnosia?
Where individuals can process and judge rudimentary aspects of objects, but can’t combine these into meaningful wholes.
What is associative agnosia?
Individuals can perceive and integrate visual information to form holistic representations, which cannot be linked to stored knowledge.
