Object recognition II

Question 1

Biedermann model

Answer 1

Proposed that we recognize objects by components, goons (geometric shapes)
critics are: how is it possible to distinguish between objects of similar shape
what are the brains mechanisms and where are the goons located?

Question 2

Biederman on viewpoint dependent recognition

Answer 2

object recognition is affected by viewpoint, research shows that objects seen in less familiar viewpoints are less recognized

Question 3

Ullmans model

Answer 3

explains how images/objects are segmented and how they are categorized based on visual representations (hierarchically organized)

Question 4

Basic features of the model

Answer 4

highly informative pictorial fragments that are hierarchal organized
a feature is informative when it reduces uncertainty of the objects category.

informative features: need distinction and frequency

Question 5

feature hierarchies

Answer 5

help to generalize objects
atomic fragments: lines, edges, corners
bottom up: incomplete information about the object
top-down segmentation: groups together different fragments and then classify it as a certain object.

Question 6

Summary of Ullmans model

Answer 6

A hierarchy of abstract features continuously extracted from examples. they are usually very class-specific, non generic, and learned based on usefulness.

pros: can extract known facts of object
cons: takes a huge amount of extracting fragments

Question 7

IT neurons

Answer 7

respond to pictorial objects preferentially but are not highly selective. Size and color can change.

Question 8

types of features

Answer 8

low: generic and correlate directly with picture
mid: allows generalization across categories
high: very specific to a class: e.g. fruits