Perception Flashcards
What is perception?
-the process of recognizing, organizing and interpreting information from senses
What is the difference between perception and sensation?
-sensation is objective and perception is subjective
What is the inverse projection problem and why is it hard for computers to observe similar objects?
-they confuse objects that share features (think of like a tennis ball and a circular plate)
-the human brain can go back into memories and figure out what it is
What is an example of a challenge of object perception?
-three shapes at different angles that are all different shapes but look like the same shape because of the way they are angled and the depth they are at. Humans can figure out these are different shapes but AI struggles to
Why do AI struggle with hidden or blurred images? Think of a hidden pair of scissors, we know what they are but the AI program may struggle. Or blurred photos of celebrities.
-because we can go into our memory but a computer struggles with that
What is viewpoint invariance?
-the ability to recognize an object from different viewpoints
What is bottom-up processing?
-starts with the sensory receptors and then works up to the brain to identify the object/scene being perceived
What is top-down processing?
-based on the knowledge we have
What are the three bottom-up approaches we discuss? (3)
-template matching theory
-feature approaches (Selfridge)
-recognition by components
What is template matching theory? Example. (2)
-we compare a stimulus with a set of templates we have in our memory and identify the object when a match is actually made
Example: the barcodes on items at the grocery template match the bar codes in the system which brings up the price
What is a huge downfall of template matching theory for humans?
-humans are able to interpret things that are scratched out or obscured or in different people’s handwriting (we essentially don’t do this)
What is the feature approaches pandemonium model by Selfridge?
-feature by feature analysis that allows for recognition, pattern recognition with “demons” processing visual information in stages.
What are the limits of the feature approaches pandemonium model?
-cannot breakdown extremely complex things or very simple things
What is recognition-by-components theory by Biederman?
-the way we perceive and recognize objects is by breaking them down into these individual 3D shapes he called geons
How many geons did Beiderman say we have?
-36
What are the strengths of the recognition-by-components theory? (4)
-viewpoint invariance
-can recognize objects despite visual noise
-can produce 306 billion objects
-the ability to recognize geons develops early in life
What are the weaknesses of the recognition- by-components theory? (2)
-can’t take a 2D photo of an object and work backward to describe a 3D object
-based solely on geons it is hard to distinguish between objects
What are the three top-down approaches? (3)
-Helmhotz’s theory of unconscious inference
-getalt principles of organization
-environmental regularities
What is speech segmentation?
-knowing when a word begins and ends in speech based on previous experience with the language, think about how you struggle to hear the separate words in german