W1: Recognising Objects Flashcards
In relation to Object Recognition, define form perception and object perception
Forms are perceived, not just as the sum of their parts, but also people process how the form is ranged in depth, which part of the form is figure or ground which is not contained in the stimulus itself
What’s the difference between a reversible and ambiguous figures
A Necker cube is a reversible figure (relating to depth), able to be perceived one way or another, but not both at the same time. The lines are neutral in regards to the configuration of depth.
Vase/faeces this relates to ground/figure issues. you can look at either the faces or vase but not both at once.
The interpretation for both images comes from you, not stimulus.
Gestalt principles of organisation
Similarity, proximity, good continuation, closure, and simplicity.
Simplicity
We tend to interpret the form in the simplest way possible.
Closure
Broken triangle is perceived as whole. We prefer to perceive complete figures.
Good continuation
If one object partially obscures another, we tend to see the rear object as continuous.
Proximity
Dots close together are perceived as groups
Similarity
Create groups of similarity (i.e. columns of same coloured dots)
How do ‘perception of features’ and ‘analysis of the object’s configuration’ interact?
Features and configuration have to be in place before we can interpret.
Perception of features is guided by configuration and analysis of configuration is guided by features (parallel process)
Perceptual constancy
we perceive constant properties of objects in the world even though sensory information changes as our viewing circumstances change. i.e. size, shape, brightness.
Describe the two processes that influence object recognition
o Bottom up: data and stimulus driven - look around us and take in the incoming data
o Top-down: concept driven, influenced by context, prior knowledge and concepts
Bottom up
data and stimulus driven - look around us and take in the incoming data
Top-down
concept driven, influenced by context, prior knowledge and concepts
How do visual features contribute to object recognition?
o The vertical lines, curves, diagonals and so on make up the visual features.
o Feature detectors: i.e. detector for the features of the letter ‘A’ which are somewhat flexible.
What are the two main factors that influence recognition?
Familiarity and recency
What is the experimental technique that we use to study this type of recognition?
Tachistoscope, or tachistoscopic presentations - Brief displays of word stimuli for controlled periods of time (ie 20ms). each stimulus is followed by a mask (hgpxt) - word recognition
Define the word superiority effect and describe the technique used to establish it
o Words are easier to perceive than isolated letters.
o Two alternative, forced choice procedure.
Why are degrees of well-formedness important in recognition?
Grammatical well-formedness helps you bootstrap what the word would have be..
What is a feature net and how does word recognition occur?
Bottom up processing: Feature detectors, letter detectors, word detectors
How do bi-gram detectors account for well-formedness?
Theory of activation level, response threshold, recency and frequency
Detectors of letter pairs.
o Step between letter and word detectors
o Well formedness will activate a bigram detector for familiar letter pairing, but not unfamiliar
How do we recover from the confusion that may occur in response to briefly presented stimuli?
o At the bigram level
o If we only have partial information about a letter all possible combinations are activated then bigram detectors select the most likely option (frequency/priming)
What role does distributed knowledge play in the feature net?
Knowledge is not locally represented it is distributed knowledge, represented in a fashion that is distributed across the network and detectable only if we consider how the entire network functions. i.e., there is no specific ‘CO’ detector site. The whole system is working.
How does the McClelland and Rumelhart (1981) model accomplish string recognition without bigram detectors?
Excitatory and inhibitory connections: connections that allow one connector to activate its neighbors. T activates t-words, and inhibits other detectors.
Visual processing is bi-direction.
How do people recognise objects in the recognition by components (RBC) model?
o Geons (geometric ions) simple shapes such as cylinders cones and blocks. o Using a hierarchy of detectors. The lowest level are feature detectors, ie edges curves, vertices. These activate geon detectors which then activate higher level detectors sensitive to higher combinations of geons. Viewpoint independent. Ie the back of a cat is still recognized as a cat.
Why is recognition viewpoint dependent in the recognition via multiple views approach to object recognition?
o Recognition requires mental rotation. Some viewpoints are slower than others. The speed of recognition will be viewpoint-dependent.
o A bottom up hierarchy of detectors, becoming increasingly complex up to the whole object.
Why do we think that there is a special recognition system for faces?
o Prosopagnosia seems to imply the existence of special neural structures involved almost exclusively in the recognition of faces.
o Facial recognition is a different process to other recognition thingys.
What is the evidence for holistic processing of faces?
Composite face recognition. Difficult if faces are properly aligned.
Give an example demonstrating how our knowledge about the world influences our object recognition
Top down is context and prior knowledge. Ie words are easier to recognise if you see them in a sentence,
The Necker cube demonstrates which principle of perceptual organisation?
If the input is ambiguous, the image can be interpreted in different ways at different times
What best illustrates the effect that Gestalt principles have on perception?
“Go beyond the information given.”
Jenna sees a picture of a dog standing in front of a tree. The dog is blocking part of Jenna’s view, so that she cannot see a portion of the tree trunk. Jenna does, however, perceive the tree to have an intact, continuous trunk. Jenna’s perception reminds us that:
People generally “fill in” missing perceptual information, guided by the Gestalt principles
Despite the fact that sensory stimuli can change from moment to moment, we perceive the details (colour, shape, etc.) of an image to be stable because of:
Constancy
What underlies the achievement of perceptual constancy?
Unconscious inference
What best describes Visual illusions?
Cognitive principles that generally help us can cause illusions in some cases.
What sort of processing is driven primarily by factors in the environment or in the stimulus itself?
Bottom Up.
Which task is likely to produce the longest reaction time in an experiment requiring participants to quickly determine if a target is present?
Identifying a red vertical line in a field of red horizontal and blue vertical lines.
What does a Tachistoscope device do?
Display stimuli for precisely controlled exposure times.
What is an example of word superiority effect?
HAFE > HZYE
participants have been shown non-word letter strings, presented very briefly. When asked to identify these strings, participants tend to make specific kinds of errors. How would these errors be best described?
They tend to misidentify strange letter combinations as more-common letter combinations.
Participants’ recognition thresholds are…
lower for frequently seen words.
We sometimes encounter ambiguous letters when reading handwritten words, but we can still interpret the words. For example, the same shape can be interpreted as an A in CAT but an H in THE. At what level of analysis does the feature net resolve this issue?
At the word level.
Knowledge of some sorts is likely to be represented by a broad pattern of activation spread across a network which represents…
Distributed representation
Mistakes in word recognition occur within a feature net model of recognition. One reason is feature net encourages ________ over ________.
Efficiency over accuracy
What is the term “geons” short for?
Geometric ions
What best describes viewpoint-dependent object recognition?
The perceiver must match the current view of an object with a view of the object stored in memory, often using the process of rotation.
Brain damage identified as Prosopagnosia disables what?
Inability to recognise faces.
Facial recognition depends on the configuration and spacing of the features, which reflects what type of processing?
Holistic
Priming effects do what?
- Change in response to a stimulus caused by exposure to an identical, similar, or related stimulus;
- Impact the words we perceive;
- Can meaningfully impact our understanding of situations.
What are bigram detectors?
Detectors of letter pairs.
What is repetition priming?
When the test stimulus is the same as or resembles the priming stimulus
McClelland and Rumelhart Model
A model referring to excitatory and inhibitory connectors involved in pattern recognition.
recognition by components theory
a specific view of an object can be represented as an arrangement of simple 3-D shapes called geons (geometric icons).
