VisualBufferStd Class Reference

An integrative visual buffer, std implementation. More...

#include <Neuro/VisualBuffer.H>

Inheritance diagram for VisualBufferStd:

[legend]

Collaboration diagram for VisualBufferStd:

[legend]

List of all members.

Public Member Functions
	VisualBufferStd (OptionManager &mgr, const std::string &descrName="Visual Buffer", const std::string &tagName="VisualBuffer")
	Constructor.
virtual	~VisualBufferStd ()
	Destructor.
Protected Member Functions
	SIMCALLBACK_DECLARE (VisualBufferStd, SimEventAttentionGuidanceMapOutput)
	Callback for every agm.
	SIMCALLBACK_DECLARE (VisualBufferStd, SimEventRetinaImage)
	Callback for every agm.
void	input (const WTAwinner &win, const Image< float > &sm, const Image< byte > &objmask)
	New input.
Image< byte >	getSaliencyMask () const
	Get saliency mask.
Image< float >	getBuffer () const
	Get current buffer.
void	evolve (SimEventQueue &q)
	evolve our internal dynamics
void	inhibit (const Point2D< int > &loc)
	inhibit a location, given in retinotopic/retinal-scale coords
void	inhibit (const Image< float > &mask)
	inhibit a location, using a mask of same dims/scale as the buffer
Point2D< int >	findMostInterestingTarget (const Point2D< int > &p)
	Decide on most interesting target location given current eye position.
Point2D< int >	findMostInterestingTargetLocMax (const Point2D< int > &p)
	Decide on most interesting target location given current eye position.
Point2D< int >	retinalToBuffer (const Point2D< int > &p) const
	transform coord from retinotopic/retinal-scale to world-centered/sm-scale
Point2D< int >	bufferToRetinal (const Point2D< int > &p) const
	transform coord from world-centered/sm-scale to retinotopic/retinal-scale
bool	isObjectBased () const
	are we object-based?
void	start1 ()
	get started (see ModelComponent.H)
Protected Attributes
OModelParam< int >	itsFOAradius
	FOA radius.
OModelParam< bool >	itsIgnoreBoring
	ignore boring attention shifts
OModelParam< bool >	itsObjectBased
	true if doing object-based
OModelParam< Dims >	itsBufferDims
	dims of our internal buffer
OModelParam< LevelSpec >	itsLevelSpec
	levelspec determines buffer scale
OModelParam< Dims >	itsInputDims
	input image dims
OModelParam< SimTime >	itsTimePeriod
	period at which to apply interaction
OModelParam< float >	itsDecayFactor
	temporal decay factor
OModelParam< MaxNormType >	itsNormType
	maxnorm type for our internals

Detailed Description

An integrative visual buffer, std implementation.

A world-centered Visual Buffer to accumulate interesting locations (each time a covert attention shift occurs), with internal competitive dynamics within the buffer.

Definition at line 132 of file VisualBuffer.H.

Constructor & Destructor Documentation

VisualBufferStd::VisualBufferStd	(	OptionManager &	mgr,
		const std::string &	descrName = `"Visual Buffer"`,
		const std::string &	tagName = `"VisualBuffer"`
	)

Constructor.

Definition at line 136 of file VisualBuffer.C.

VisualBufferStd::~VisualBufferStd ( ) [virtual]

Destructor.

Definition at line 170 of file VisualBuffer.C.

Member Function Documentation

Point2D< int > VisualBufferStd::bufferToRetinal ( const Point2D< int > & p ) const [protected]

transform coord from world-centered/sm-scale to retinotopic/retinal-scale

Definition at line 387 of file VisualBuffer.C.

References Image< T >::getDims(), OModelParam< T >::getVal(), itsLevelSpec, LevelSpec::mapLevel(), and visualBufferToRetinal().

void VisualBufferStd::evolve ( SimEventQueue & q ) [protected]

evolve our internal dynamics

Definition at line 290 of file VisualBuffer.C.

References OModelParam< T >::getVal(), itsTimePeriod, and SimEventQueue::now().

Point2D< int > VisualBufferStd::findMostInterestingTarget ( const Point2D< int > & p ) [protected]

Decide on most interesting target location given current eye position.

All coordinates are in world-centered/sm-scale. The algorithm used here is inspired from the area activation model; we threshold out regions lower than a fraction of the maximum activation, then aim for the centroid of the region that is closest to current eye position.

Definition at line 317 of file VisualBuffer.C.

References centroid(), chamfer34(), findMin(), flood(), Image< T >::getDims(), getMinMax(), inplaceReplaceVal(), makeBinary(), and ZEROS.

Point2D< int > VisualBufferStd::findMostInterestingTargetLocMax ( const Point2D< int > & p ) [protected]

Decide on most interesting target location given current eye position.

All coordinates are in world-centered/sm-scale. The algorithm used here will aim for the local max above a threshold that is closest to current eye position.

Definition at line 345 of file VisualBuffer.C.

References Point2D< T >::distance(), Image< T >::getHeight(), getMinMax(), Image< T >::getVal(), Image< T >::getWidth(), Point2D< T >::i, and lowPass9().

Image< float > VisualBufferStd::getBuffer ( ) const [protected]

Get current buffer.

Definition at line 377 of file VisualBuffer.C.

Image< byte > VisualBufferStd::getSaliencyMask ( ) const [protected]

Get saliency mask.

This is mostly for display purposes. The mask is the one that was created at the last input() and has the same size/scale as the saliency map and is in retinal coordinates

Definition at line 286 of file VisualBuffer.C.

void VisualBufferStd::inhibit ( const Image< float > & mask ) [protected]

inhibit a location, using a mask of same dims/scale as the buffer

Parameters:

mask

should have same dims as the buffer, and values between 0.0 (no inhibition) and 1.0 (full inhibition).

Definition at line 310 of file VisualBuffer.C.

References binaryReverse().

void VisualBufferStd::inhibit ( const Point2D< int > & loc ) [protected]

inhibit a location, given in retinotopic/retinal-scale coords

A disk of radius FOAradius will be used for inhibition

Definition at line 302 of file VisualBuffer.C.

References drawDisk(), Image< T >::getDims(), OModelParam< T >::getVal(), itsFOAradius, itsLevelSpec, LevelSpec::mapLevel(), and ZEROS.

void VisualBufferStd::input	(	const WTAwinner &	win,
		const Image< float > &	sm,
		const Image< byte > &	objmask
	)			`[protected]`

New input.

Beware that the inputs here are matched to what Brain naturally provides, which is

Parameters:

	win	the coordinates of an atteention shift at retinal resolution
	sm	the saliency map at the maplevel resolution
	objmask	a mask with (if initialized) the shape of the attended object at retinal resolution.

Definition at line 211 of file VisualBuffer.C.

References binaryReverse(), WTAwinner::boring, chamfer34(), Image< T >::clear(), drawDisk(), Image< T >::getDims(), OModelParam< T >::getVal(), Dims::h(), Image< T >::initialized(), inplaceLowThresh(), itsFOAradius, itsIgnoreBoring, itsInputDims, itsObjectBased, WTAwinner::p, pasteImage(), rescale(), Image< T >::resize(), retinalToBuffer(), squash(), takeMax(), Dims::w(), and ZEROS.