Chapter 5 - Perceiving Objects and Scenes Flashcards
1- Computer vision. Why is it hard to design a perceiving machine?
- What can computers do?
- Computer vision
- Object recognition
-Detection of objects in an image and then
matching those objects to existing, stored
representations of what those objects are to
create a scene.
-Bat + boy = boy holding baseball bat. - Other computer systems
-Learn how to recognize objects and determine not a description of a
scene, but rather, the precise locations of objects in that scene. - Autonomous vehicles
- Require fast and precise identification of objects in order to smoothly navigate the environment.
- Cellphones
-Rely on object recognition
-Recognize faces across
different angles and lighting
conditions to unlock your
device. - Computers becoming more
accurate - In some situations, their
object detection
performance sometimes
matches or even exceeds
that of humans. - However, they often fall short is in
identifying objects under
degraded conditions.
Why Is It Hard to Design a Perceiving Machine?
The stimulus on the receptors is ambiguous.
* Inverse projection problem: an image on the retina can be caused by an infinite number of objects.
Objects can be hidden or blurred.
* Occlusions are common in the environment.
Objects look different from different viewpoints.
* Viewpoint invariance: the ability to recognize an object regardless of the viewpoint.
* This is a difficult task for computers to perform.
2- Perceptual organization.
Approach established by Wundt (late 1800s)
* States that perceptions are created by combining
elements called sensations.
-Structuralism could not explain apparent movement.
* Stimulated the founding of Gestalt psychology in the
1920s by Wertheimer, Koffka, and Kohler.
* The whole differs from the sum of its parts.
-Perception is not built up from sensations but is a
result of perceptual organization.
3- The Gestalt approach to perceptual grouping.
Structuralism: distinguished between sensations and perceptions
Apparent movement: illusion of movement
Illusory contour: appear real but have physical edge
Principles of perceptual organization:
* Good continuation: connected points resulting in straight or smooth curves belong together.
-Lines are seen as following
the smoothest path.
* Pragnanz: every stimulus is seen as simply as possible.
* Similarity: similar things are
grouped together
* Proximity: things that are near to each other are grouped together.
* Common fate: things moving in the same direction are grouped together.
* Common region: elements in the same region tend to be grouped together.
* Uniform connectedness:
connected region of visual
properties are perceived as single unit.
(see others)
Perceptual Segregation
Figure-ground segregation: determining what part of the
environment is the figure, so that it “stands out” from the
background.
* Properties of figure and ground:
-The figure is more “thinglike” and more memorable than the ground.
-The figure is seen in front of the ground.
- The ground is more uniform and extends behind figure.
- The contour separating figure from the ground belongs to the figure (border ownership).
* Figural cue proposed by the Gestalt psychologists:
-Areas lower in the field of view are more likely to be perceived as figure.
4- The Role of Perceptual
Principles and Experience in
Determining Which
Area Is Figure vs. Meaningfulness
- Gestalt psychologists’
emphasis on perceptual
principles led them to
minimize the role of a
person’s past experiences in
determining perception.
Meaningfulness
* Bradley Gibson and Mary Peterson (1994)
* experiment that argued against this idea by
showing that figure–ground formation can be affected by the meaningfulness of a stimulus (ex: looks like a person)
5- Recognition by components
- Recognition by components (RBC) theory.
- Objects are comprised of individual geometric components called geons,
- Geons:
- Three-dimensional shapes, like pyra-mids, cubes, and cylinders.
- 36 different geons from which most objects we encounter can be assembled and recognized.
- Shortcomings
- Many aspects of object perception that the RBC theory could
not explain.
-Grouping or organization like the Gestalt principles do. - some objects can’t be represented by assemblies of geons (like clouds in the sky that typically don’t have geometric components).
- The RBC theory also doesn’t allow for distinguishing
between objects within a given category, such as two
different types of coffee mugs or species of birds that
might be composed of the same basic shapes.
6- Perceiving scenes and objects in scenes. Perceiving the gist of a scene.
- Scene
- (1) background elements,
-(2) multiple objects that are organized in a meaningful
way relative to each other and the background. - Objects
-compact and acted upon. - Scenes
-Extended in space and are acted within.
Perceiving the Gist of a Scene
* Mary Potter (1976)
* showed observers a target picture and then asked them to indicate whether they saw that picture as they viewed a
sequence of 16 rapidly presented pictures.
* Her observers could do this with almost 100 percent
accuracy even when the pictures were flashed for only 250 ms (ms 5 milliseconds; 250 ms 5 1/4 second).
* Even when the target picture was only specified by a written description, such as “girl clapping,” observers
achieved an accuracy of almost 90 percent
Li Fei-Fei’s experiment (2004)
7- What enables observers to perceive the gist of a scene so rapidly?
- Aude Oliva and Antonio Torralba (2001, 2006)
- Global image features
- Can be perceived rapidly and are associated with specific types of scenes.
- Degree of naturalness
- Degree of openness.
- Degree of roughness.
- Degree of expansion.
- Color.
- Past experiences in perceiving properties of the environment
- Blue is associated with open sky.
- landscapes are often green and smooth.
- verticals and horizontals are associated with buildings.
Regularities in the
Environment: Information
for Perceiving
* Physical regularities
* Regularly occurring physical properties of the environment.
* There are more vertical and
horizontal orientations in the
environment than oblique (angled) orientations.
* This occurs in human-made
environments
* buildings contain many
horizontals and verticals
* in natural environments (trees and plants are more likely to be vertical or horizontal than slanted).
Indentation vs. bump effect (light comes from above)
Semantic regularities
* Semantics
-The meaning of a scene.
-Scene schema.
8- The role of inference in perception
- Helmholtz’s Theory of
Unconscious Inference - Likelihood principle
-We perceive the object that
is most likely to have caused
the pattern of stimuli we
have received. - Unconscious inference
-Our perceptions are the
result of unconscious
assumptions, or inferences,
that we make about the
environment. - Bayesian inference
- We perceive what is
most likely to have
created the stimulation
we have received in
terms of probabilities.
(ex: coughing… from cold, lung disease or heart-burn?)
9- How the brain implements prediction.
- Predictive coding
- A theory that describes how
the brain uses our past
experiences—or our “priors,”
as Bayes put it—to predict
what we will perceive. - Prediction error signal
10- Connecting Neural Activity and Object/Scene
Perception
- Brain Responses to Objects and Faces
- lateral occipital complex (LOC)
- Active when the person views any kind of
object—such as an ani-mal, face, house, or tool—but not when they view a texture,
or an object with the parts scrambled. - builds upon the processing that took place in lower-level visual regions, like V1 where
the neurons responded to simple lines and
edges.
11- The Neural Correlates of Face Perception
- Nancy Kanwisher
- Fusiform face area (FFA)
- fMRI to determine brain activity in response to pictures of faces and other objects such as household objects, houses, and
hands. - Subtracted the response to the other objects from the response to the faces.
- Prosopagnosia
- (Greek for “prosopon” = “face” and
“agnosia” = “not knowing”) - Difficulty recognizing the faces of familiar people.
12- Neural Representation of Other Categories of Objects
- Extrastriate body area (EBA)
- Activated by pictures of
bodies and parts of bodies - Alex Huth and coworkers (2012).
- Participants viewed 2 hours of film clips while in a brain
scanner. - Analyze how individual brain areas were activated by
different objects and actions in the films. - Parahippocampal place area (PPA) (parahippocampal cortex (PHC)).
- Spatial layout hypothesis.
13- The Relationship
Between Perception and
Brain Activity
- Frank Tong and coworkers (1998)
- Binocular rivalry.
- When the observers perceived the house,
activity occurred in the parahippocampal
place area (PPA) in the left and right hemispheres (red ellipses). - When observers perceived the face, activity occurred in the fusiform face area (FFA) in the left hemisphere (green ellipse).
