Lecture 29 Flashcards
gestalt principles tend to operate on a ____ basis
committee
surroundedness
if one region is entirely surrounded by another it is likely the figure
size
smaller region likely figure`
symmetry
symmetrical region is more likely to be seen as a figure
parallelism
regions with parallel contours are more likely to be seen as a figure
relative motion
details move relative to an edge can help determine which portion is figure and which is ground
bayes theorem
mathematical model that enables us to calcualte the probability (P) that the world is in a particular state (A) given a particular observation (O)
P(A/O) = P(A) x P(O/A)/P(O)
object recognition is more of a(n) __________ process
anterior
parahippocampal place area
activated mainly by pictures of places
fusiform face area
activated by images of faces
extrastriate body area
activated mainly by body parts other than the face
deep neural networks
a type of machine learning in artificial intelligence in which a computer is programmed to learn something
process of teaching deep neural networks
- In the first layer of the DNN, a set of features is extracted from the image (think: simple cells)
- Next, information is pooled (think: complex cells)
- These operations create a new image from which the next layer of the DNN will extract features
- These features are pooled… and so on
- The top layer has specific neurons for each category you are trying to identify (akin to “grandmother cell” coding)
how does this network “learn”
- Connections between the layers in the DNN are governed by weights, which are initially random
- Before training, the categories fit random loose connection
- Showing the network many images and correcting weights based on computer performances -> training the network
- Lots of artificial intelligence potential with application (i.e. medical diagnosis)
- A properly trained network that is fed a new image should be able to correctly categorize the image
are faces processed differently from other objects
yes
