Topic 4: Perceiving Objects and Scenes Flashcards
The importance of knowledge for perception is highlighted by what difference?
The difference between human perception and computers/AI
What are five situations/domains in which computers often struggle with object recognition?
When different objects are similar sizes/shapes
Degraded conditions
Unexpected situation
Hidden objects
Viewpoint invariance
What are four difficulties in perception for humans?
The image on the retina is ambiguous, such that multiple objects can create a similar or the exact same image on the retina.
The image on the retina also does not always accurately reflect what is really “out there”
Often make “assumptions” or “predictions” about what is the most likely cause of what we are perceiving.
Viewpoint also plays a role and can impact the image on the retina as well.
Object properties don’t change when the object turns
Viewpoint invariance
Process by which elements are grouped & segregated to create a perception.
Perceptual organization
The perspective that much like how individual atoms come together to form complete chemical structures, simple stimuli combine to form complete perceptions
Structuralism
Approach that rejected the idea that perceptions were formed by “adding up” individual stimuli
Gestalt approach
An illusion where movement is perceived when nothing is moving
Apparent motion
What are the seven Gestalt principles of perceptual organization?
Principle of good continuation
Principle of pragnanz/good figure/simplicity
Principle of similarity
Principle of proximity
Principle of common fate
Principle of common region
Principle of uniform connectedness
Points that, when connected result in straight or smoothly curving lines are seen as belonging together, and the lines tend to be seen in such a way as to follow the smoothest path
Principle of good continuation
Every stimulus pattern is seen in such a way that the resulting structure is as simple as possible
Principle of pragnanz/good figure/simplicity
Similar things appear to be grouped together.
Principle of similarity
Things that are near each other appear to be grouped together.
Principle of proximity
Things that move in the same direction appear to be grouped together
Principle of common fate
Things that are within the same region of space appear to be grouped together; Can override proximity
Principle of common region
Connected objects/regions with similar visual properties are perceived as grouped together
Principle of uniform connectedness
The perceptual separation of an object from its background
Figure-ground segregation
A figure-ground pattern that perceptually reverses as it is viewed, such that the “figure” and the “ground” are often ambiguous
Reversible figure-ground
What are the four “rules” for deciding what is the figure and what is the ground?
The figure is seen as being in front of the ground, and the ground is seen as extending behind the figure
The shared border is perceived to belong to the figure, and not the ground (see the previous image)
Things that are lower in our field of view are more likely to be perceived as the figure.
Things that align with expectations/previous knowledge are more likely to be the figure.
Proposes that objects are comprised of individual geometric components called geons, and we recognize objects based on their arrangement
Recognition by Components (RBC) theory
Three-dimensional shapes that are the building blocks of objects and can be arranged in different ways to form different objects (RBC)
Geons
Name one advantage and two disadvantages to the RBC approach
Advantage: Is unaffected by viewpoint invariance
Disadvantages: Some objects simply can’t be represented easily (abtract objects), can’t easily distinguish within the same category
An environment that contains background elements alongside multiple objects that exist relative to each other
Scene
Regularly occurring physical properties of the environment
Physical regularities
Name two key physical regularities
Horizontals and verticals
Light-from-above assumption
Characteristics associated with functions carried out in different types of scenes
Semantic regularities
We look for (and often find) what we expect to find in a scene based on knowledge
Scene schema
What do scene schemas do in relation to perception?
Assigning a meaning to a scene can help us determine what an object probably is within that scene, even when we can’t necessarily perceive the scene very well
