Perception

Goal: Perception=for action

Direct Perception

What?

Bottom up

Ecologist: Most info lost in Laboratory Studies

Visual Symstem developed in real world

Real World has enough clues

Affordance

Objects guide action: make evident what can be done with them

No memory involved

No processing

Evaluation

Strong theory: explains a lot

Need for realistic / non lab investigations

Critque

We do make assumptions

Cannot explain phenomena

Size constancy

Image on the retina gets bigger but doesnt jump at us

Visual illusions

Kanizsas Triangle

Neckar Cube

Doesn't explain how it works

Evidence

Work with pilots

Training with 2D little impact

pilots undistinguishable with 2D Training material

How?

Because of poss. ambiguity we Search world for invariant clues

Depth Clues

Vergance to Motor

Texture Gradients

Faster moving= closer (motion parallax)

Far colours more bluish

Relation to Horizon

occlusion: close cows cover up far cows

Schneiders Hamsters

Pattern recog. vs. orientation in space

Carlton / Binsted

Learning example

Intital Learning: ventral

More conscious processing

Once Learned: dorsal

More Motor action

Normans: Dual Proces Approach

More interconnected than independent

Evidence for 2 routes

Patient DF (Milner and Goodale

Not recog objects / not know size difference

but grab right size with her fingers

Ventral

What= Gregory

Drives Object recogntion

Dorsal (Fast)

Where= Gibson

Drives behaviour

Most stable representationwill occur: Gestaltfest

What?

Rules built in

Marr: for fore and background

Subtopic

How?

Not fully developed: No "How"

Critique

Only explains simple phenomena

Goal: Perception=for Recognition

How?

Top Down

Constantly generate hyptothesis

Prior knowledge= important

Law of Prägnanz (best gestalt will appear)

Why?

Incomplete info: Perception under uncertainty

Ambigous Clues

Critque

Makes use of visual illusion

Ames Room: We know there is no giant/dwarf

What?

Mostly Gibsonian with Gestaltist

How?

Bottom up

Processing

Algorithms

Evaluation:

Marr works but doesn't have to be like that

has been implemented

Enourmously important theory for cog psy

1. Gray level description

Critique: Colour is important clue

Marr: different module

Why Blurring

Edges: Large changes of intensity

Random fluctionations: Small Changes

2. Raw Primal(shapes)

Points are connected

Blobs (enclosed arae)

Bars (pararllel lines)

Edge: Change in intensity

Termination: Discontinuity

3. Full Primal

Objects are grouped

Gestalt Principles

4. 2,5 D

Vectors: Distances/ orientation are determined

Evidence: Motor and Texture processed sepeartly

Uses Gibson (other modules):

Stereopsis

Texture

We can move around

vs. 3D

We can recognize from different angles

The Ship

imporverished stimuli are easier recognized one we know what the object is

Knowledge is involvedWe produce Perceptional Hypothesis

Gregory

Strong top down

Size constancy

Strong evidence against direct perception

Kanizsas Triangle

See Triangle although there is none

Necker cube

We constantly build a new perceptual hypo

Most likely one wins

Mask of Hor

Hypo: it must be face

Wrong perception

Top Down Knowlege

no face, its hollow

Direct perception= no processing

Object + Environmental clues are enough to create percept

Ames room

We know there are no giants

Yet we see one

Penrose Triangle

Hyp: it must be a 3D triangle

Top-Down:

But its 2D

Evaluation

Knowledge sometimes but not allways involved

So perception depends on the task

Use of visual illusion can be used for both positions