Perceiving objects Flashcards
Top-down (conceptually driven)
perception is influenced by our prior knowledge, memories and experiences
Bottom-up (data-driven)
recognize patterns by analyzing sensory input step by step
Problems with direct perception
- the stimulus on the receptors is ambiguous
- Objects can be hidden or blurred –> one object can be in front of the other or you may not have a clear view (visual occlusions)
- Objects look different from different viewpoints
RBC Theory
RBC helps compute vision, all objects are made up on geons (3D shapes) – perception is the looking at an object and breaking it apart by the geons. —
assumes that object recognition involves the identification of simple geometric shapes, known as geons, and their spatial arrangement within an object
Geons
Geons are viewpoints invariant and show discriminability (we can recognize basic shapes [geons] form any angle and tell them apart) because they each have a unique set of non-accidental properties
Non accidental properties
characterics of geons that belong to geons, don’t see them by accident, belong to object
Examples against RBC
- Holistic processing: Humans often recognize objects holistically rather than by isolating and analyzing individual components
- Context influence: Context and global features impact object recognition, challenging the idea that recognition is solely based on components
Canonical perspective
an object is the preferred perspective for viewing that object (ex. Drawing a car, but not the bottom of the car)
Changes in object’s perspective caused changes to:
- “Goodness” rating of object photographs (aesthetics)
- Response time/error rates while identifying objects
- Formation and descriptions of mental images of objects
What causes a perspective to be canonical or noncanonical
- Bottom-up (informative perspective)
The neurons associated with the recognition of an object become tuned to the features present in the most recognizable (informative) view of the object
- Top-down (experienced perspective)
Canonical viewpoints for novel objects form only after repeated viewings at a specific orientation
Gestalt psychology
- Law of proximity
- Law of similarity
- Law of good continuation
- Law of closure
- Law of common fate → in motion, but end up in the same location, group together
- Law of common region → exists within one region, grouped together
- Law of uniform connectedness →
Figure-ground segregation
- Figure appears closer to you
- Contour ‘belongs to’ the figure
- Ground appears farther away and extends behind the figure
- Size plays a role (smaller items tend to be seen as the figure)
- Orientation: prefer vertical than horizontal
a. Contrast
b. Closure
c. Symmetry
d. Convexity
Law of perceptual organization
They are ceteris paribus rules: they hold only when all else is equal
We can use these rules to predict what will be perceived based on one law at a time – it is hard to predict the outcome of combining laws
Role of experience
If things have been associated in prior viewings, they will be grouped together in the future
Our perceptual system typically won’t make consciousness resulting from an accidental viewpoint
Scene perception
Top down processing is particularly evident during scene perception
We use semantic regularities to help identify objects in a scene
Gist processing of a visual scene is fast, automatic, and relies on past experience
We can use physical regularities ot help perceive global image features of a scene
a. Degree of naturalness
b. Degree of openness
c. Degree of roughness (texture/details/lines and edges)
d. Degree of expansion
e. Colour
