Lecture 13 Flashcards
Discuss our object perception ability
We can recognise different classes of things, for example, cartoons, pictures, real life etc.
We recognise the same object from different angles and sometimes when part of it is occluded.
We can recognise objects extremely rapidly.
It’s a very complex process to recognise an object.
When objects overlap, most of the time we can tell where one ends and the other starts.
Objects can vary enormously but we can still recognise them as the same object. e.g. types of chairs.
We can also recognise objects from different distances.
Discuss bottom up vs top down object recognition
We can recognise large scale shapes as quickly as small details whereas computers use a bottom-up approach as they begin with pixels and build up an image. Some research suggests we see the bigger picture/context first (top down) whereas other research suggests that we can choose which we see first.
Discuss constancies
When something is the same or unchanging. A colour constancy is when we can perceive an object as the same colour even when in different illuminations, however, we can sometimes be incorrect here, allowing illusions to occur. For example, we can think two objects are different colours as a result of a shadow and the context around it, however, they are both the same shade.
Discuss object perception and context
Sometimes recognition can require context in order to perceive it. For example, if a word is slightly obscured, then we can look at the sentence around it, allowing us to perceive the word.
Discuss the levels of recognising an object
First, we see the basic features (edge perception), then we perceive organisation (working out which features belong together, Gestalt principle), then shape, followed by comparing this shape to your memory of shapes and then finally we make a decision as to whether the object is familiar. If so, we use our memory to gather information about the object.
What is simultanagnosia?
The disorder where you can’t perceive how parts fit together, you can see separate objects but not understand the whole scene.
Discuss the level of processing: perceiving shape
There can be issues with this because there’s two levels of processing here; perceiving figure (central) or perceiving ground (background) which can contradict each other. For example, the illusion where there’s a white vase with a black background poo or you could view it as two faces looking at each other with a white background. We usually perceive things globally first but sometimes when we look at it locally, we can see problems with our initial viewpoint. For example, the impossible fork, it seems normal at first but when looking at the fork locally, we can see the problem with it. We try to work it out by attempting to separate figure and ground, however they merge.
When finally deciding the object that you are perceiving, there are different models of recognition. Discuss these.
Template matching: This process involves comparing the whole object to the stored representations in your brain. This is similar to a computer recognising numbers on cheques. However, it’s poo implausible for humans as there are so many variations of a singular object that our brain wouldn’t be able to hold that much information. Furthermore, features can be grouped together which adds another layer of recognition.
Feature analysis: This involves breaking down the shape into critical features. These features are recognised and assembled into objects/shapes which are then compared to mental templates. This can be described metaphorically with the pandemonium model. This model states that their are two ‘demons’: the feature and the cognitive demons. Feature demons detect basic aspects like angles and horizontal lines whereas cognitive demons respond when particular configurations of these features are present. However, it doesn’t account for the context as it’s a reductionist approach, it doesn’t account for 3D objects or different versions of the same object and it doesn’t account for learning new objects. On the plus side, it is a stronger approach compared to template matching as it’s more organised and there’s less stored representations needed. Also, they have found physiological evidence for this argument.
Marr and Nishihara’s model: There are 3 different levels: edge image (2D description of main light intensity changes, allowing perception of edges, contours etc.), 2 1/2 D sketch (depth and orientation of images including shading, texture etc.) and 3D model representation (self explanatory. So firstly, you identify the main axis of the object, then the axes of smaller portions, then the context and overall image. This is an example of top down processing as it uses context. However, there is no strong data to support this.
Biederman’s recognition by components: This believes that perception involves recognising geometric items in an object. This allows viewpoint invariance as you can recognise geometric shapes from any angle. This is a bottom up approach as it doesn’t use context. However, if you rotate unfamiliar objects then it takes you longer to recognise the object, which shows that object recognition can be viewpoint dependent.
Object recognition depends on sensory information and prior expectation. Context and motion aren’t very well captured into models of recognition.
Discuss motion perception
We are able to pick up changes in a stimulus over short periods of time as well as being able to recognise the same stimulus in different positions. This allows the phi phenomenon to occur, when we see a series of still images and we visualise a moving image, for example pictures of a horse running. This also allows us to see point light displays which is when a number of lights perform a motion and we see it as an organism (e.g. a human walking). Apparent motion and real motion are similar so we can think of motion as the integration of discrete views. poo
What is Pareidolia?
When people see a pattern in things when none exists, for example perceiving faces on inanimate objects like marks on a wall.
Is face recognition special?
Yes, face recognition can be argued to be seen in neonates. Nakono found that neonates orientate themselves towards upright face stimuli, this then disappears for a few months before returning. It’s also special because of configural processing; we process the face as a whole via spatial relations between facial features, which can make us ignore individual similarities among different faces but when parts of the face are unaligned, it becomes evident (Young et al.). Additionally there is the Thatcher illusion when the eyes and mouth are upside down but the face appears normal when upside down. There’s also multiple brain areas for face recognition like the fusiform face area.
Discuss some disorders for face and object recognition
Object agnostics have trouble recognising objects and face agnostics have the same issue with faces. This can be caused when one becomes a split brain patient as you are unable to analyse the stimuli you are receiving.
Prosopagnosia is a condition that makes face recognition harder; ‘you are blind to faces’, this can be caused by brain damage.
Capgras syndrome is a disorder where you have a delusion that someone you know has been replaced by an identical imposter. This could be because if the emotional pathway for face recognition is damaged then you would recognise someone but not feel them emotionally.
Discuss Bruce and Young’s model about face recognition
They found that there are many levels of processing. For example there is expression analysis (reading emotions from facial expressions), face recognition units (structural information about known faces) and so on.
What are some of the unknown things about facial recognition?
There is one major issue with configural processing because if it’s all about shape then it doesn’t explain why faces are harder to recognise when in negative. Also, distorting a face doesn’t have much effect on facial recognition either. We also recognise and process familiar faces differently to unknown faces which are harder to process. We can recognise familiar people at a further distance and with poorer image quality but it is unknown why. All of these things show how context is also extremely important when recognising faces
