Recognise visual inputs, match exact pattern of light on retina with previously remembered patterns
2 of 10
Feature detector models?
feature detector neurons:Code for specific objects
3 of 10
Structural description model?
Goals of vision, describe object unambiguously
4 of 10
What do we want?
A system that is invariant to transformations in view point and illumination
5 of 10
Marr and Nishihara's criteria?
Accessibility, scope, uniqueness, stability and sensitivity
6 of 10
Which co-ordinate system should this representation use?
Viewer centred, object centred
7 of 10
What are Geons?
Geons have properties that are retained in any 2D projection: Co-linearity, symmetry, parallelism and co-termination
8 of 10
How many geons are there?
Less than 36
9 of 10
View dependent model
Brute association of arbitary image with object, cordinate system is viewer based, primitives are sub regions of the image, not whole image as in early template matching models
10 of 10
Other cards in this set
Card 2
Front
Template matching model?
Back
Recognise visual inputs, match exact pattern of light on retina with previously remembered patterns
Comments
No comments have yet been made