featureClusterVision< FLOAT > Class Template Reference

main class for VFAT. This combines all the methods etc. More...

#include <VFAT/featureClusterVision.H>

Inheritance diagram for featureClusterVision< FLOAT >:

[legend]

Collaboration diagram for featureClusterVision< FLOAT >:

[legend]

List of all members.

Public Member Functions
	featureClusterVision (OptionManager &mgr, const std::string &descrName, const std::string &tagName, nub::soft_ref< StdBrain > &_brain, nub::soft_ref< InputFrameSeries > &_ifs, const std::string &extraArg0)
	featureClusterVision (OptionManager &mgr, const std::string &descrName, const std::string &tagName, Image< FLOAT > salMap, typename std::vector< Image< FLOAT > > cmaps, nub::soft_ref< InputFrameSeries > &_ifs, const std::string &extraArg0)
	fCV without a std brain, user must supply saliency stuff externally
void	fCVsetUpfCV (OptionManager &mgr, const std::string &descrName, const std::string &tagName, nub::soft_ref< InputFrameSeries > &_ifs, const std::string &extraArg0)
	called by default constructor to initialize this class
void	fCVmixChannels (std::vector< std::vector< FLOAT > > data, int ch1a, int ch1b, int ch2a, int ch2b, typename std::vector< FLOAT > outAlpha, typename std::vector< FLOAT > outBeta, typename std::vector< FLOAT > outGamma, FLOAT norm1, FLOAT norm2, int size)
	When called, this mixes the four channels together into a single channle.
void	fCVmixChannels (Image< FLOAT > &img0, Image< FLOAT > &img45, Image< FLOAT > &img90, Image< FLOAT > &img135, Image< FLOAT > Alpha, Image< FLOAT > Beta, Image< FLOAT > *Gamma)
	When called, this mixes the four channels together into a single channle.
void	fCVfindMixedChannels ()
	This calles mixChannels over the entire dataset.
void	fCVcheckMixing ()
	when called this extracts the features at the most salient point
void	fCVcheckICA ()
	When called this will print out an ICA/PCA representation of each channle.
void	fCVcheckICA (int channel, bool findMixed)
	When called this will print out an ICA/PCA representation of each channle.
void	fCVcheckMotionCombined (long frame)
	check the output from the combined motion channel
void	fCVfeaturesToFile (std::string fileName, bool _new)
void	fCVrunStandAloneMSBatchFilter (std::string filename)
	check stand alone feature extraction
void	fCVrunStandAloneMSBatchTest (std::string filename)
	check stand alone feature extraction
void	fCVICAfeaturesToFile (std::string fileName)
void	fCVclusterImage ()
	main method to call for most purposes
void	fCVsaccadeTest (std::string _maskFile, std::string _outFile, std::string _label, std::string _fileName)
	A test method to use with raw saccade data.
void	fCVstandAloneFeatureTest (std::string _fileName)
	test stand alone feature extraction
void	fCVgetImageBaseStats (std::string _maskFile, std::string _imageFile, std::string _outFile, std::string _label)
	fetches an images base statistics
void	fCVgetImageComplexStats (std::string _maskFile, std::string _imageFile, std::string _outFile, std::string _label)
	fetches an images base statistics
void	fCVuploadImage (Image< PixRGB< byte > > &input, std::string fileName)
	upload a raw copy of the image
void	fCVprintOutClusters ()
void	fCVprintOutCovSlices (int sizeX, int sizeY)
void	fCVprintOutBayesClass ()
void	fCVprintOutNeighborClass ()
void	fCVdumpCovMatrix (std::string fileName)
std::vector< covHolder< double > >	fCVgetCovHolders ()
	get the signature data from this class
unsigned int	fCVgetCovHolderSize ()
	how many signatures do we have
void	fCVgetClusterImages (Image< PixRGB< FLOAT > > classImage, Image< PixRGB< FLOAT > > temporalImage, Image< PixRGB< FLOAT > > targetImage, Image< FLOAT > salMap)
	get back images of clusters for analysis
void	fCVprocessOutSaccadeData (std::string maskFile, std::string outFile, std::string _label)

Detailed Description

template<class FLOAT>
class featureClusterVision< FLOAT >

main class for VFAT. This combines all the methods etc.

this class takes NPclassify2, covEstimate, Brain/saliency as well as ICA/PCA methods to create a saliency based tracking and simple identification program.

There are also many test methods for the different elements in this class

Definition at line 82 of file featureClusterVision.H.

Constructor & Destructor Documentation

template<class FLOAT >

featureClusterVision< FLOAT >::featureClusterVision	(	OptionManager &	mgr,
		const std::string &	descrName,
		const std::string &	tagName,
		Image< FLOAT > *	salMap,
		typename std::vector< Image< FLOAT > > *	cmaps,
		nub::soft_ref< InputFrameSeries > &	_ifs,
		const std::string &	extraArg0
	)			`[inline]`

fCV without a std brain, user must supply saliency stuff externally

Definition at line 128 of file featureClusterVision.C.

References featureClusterVision< FLOAT >::fCVsetUpfCV().

Member Function Documentation

template<class FLOAT>

void featureClusterVision< FLOAT >::fCVcheckICA	(	int	channel,
		bool	findMixed
	)

When called this will print out an ICA/PCA representation of each channle.

This shows the end result of the ICA/PCA process on this feature channel

template<class FLOAT>

void featureClusterVision< FLOAT >::fCVcheckICA ( )

When called this will print out an ICA/PCA representation of each channle.

This shows the end result of the ICA/PCA process on each feature channel

template<class FLOAT>

void featureClusterVision< FLOAT >::fCVcheckMixing ( )

when called this extracts the features at the most salient point

Most likely you will not need to ever call this unless you are just trying to get the most salient points plus the mixed orientations

template<class FLOAT>

void featureClusterVision< FLOAT >::fCVcheckMotionCombined ( long frame )

check the output from the combined motion channel

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVclusterImage ( ) [inline]

main method to call for most purposes

Definition at line 2992 of file featureClusterVision.C.

References featureClusterVision< FLOAT >::fCVfindMixedChannels(), Timer::get(), and Timer::reset().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVfeaturesToFile	(	std::string	fileName,
		bool	_new = `false`
	)			`[inline]`

when called this will output features to a file and quit when called this will output post ICA features to a file and quit

Definition at line 1250 of file featureClusterVision.C.

References featureClusterVision< FLOAT >::fCVfindMixedChannels().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVfindMixedChannels ( ) [inline]

This calles mixChannels over the entire dataset.

Definition at line 817 of file featureClusterVision.C.

References featureClusterVision< FLOAT >::fCVmixChannels().

Referenced by featureClusterVision< FLOAT >::fCVclusterImage(), featureClusterVision< FLOAT >::fCVfeaturesToFile(), featureClusterVision< FLOAT >::fCVsaccadeTest(), and featureClusterVision< FLOAT >::fCVstandAloneFeatureTest().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVgetClusterImages	(	Image< PixRGB< FLOAT > > *	classImage,
		Image< PixRGB< FLOAT > > *	temporalImage,
		Image< PixRGB< FLOAT > > *	targetImage,
		Image< FLOAT > *	salMap
	)			`[inline]`

get back images of clusters for analysis

Definition at line 3633 of file featureClusterVision.C.

template<class FLOAT >

std::vector< covHolder< double > > featureClusterVision< FLOAT >::fCVgetCovHolders ( ) [inline]

get the signature data from this class

Definition at line 3619 of file featureClusterVision.C.

template<class FLOAT >

unsigned int featureClusterVision< FLOAT >::fCVgetCovHolderSize ( ) [inline]

how many signatures do we have

Definition at line 3626 of file featureClusterVision.C.

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVgetImageBaseStats	(	std::string	_maskFile,
		std::string	_imageFile,
		std::string	_outFile,
		std::string	_label
	)			`[inline]`

fetches an images base statistics

Definition at line 3741 of file featureClusterVision.C.

References Image< T >::getHeight(), PixRGB< T >::getHSV(), Image< T >::getVal(), Image< T >::getWidth(), RASFMT_PNM, Raster::ReadGray(), Raster::ReadRGB(), and sqrt().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVgetImageComplexStats	(	std::string	_maskFile,
		std::string	_imageFile,
		std::string	_outFile,
		std::string	_label
	)			`[inline]`

fetches an images base statistics

Definition at line 3914 of file featureClusterVision.C.

References abs(), Image< T >::beginw(), CONV_BOUNDARY_ZERO, convolve(), downscaleFancy(), Image< T >::endw(), featureClusterVision< FLOAT >::fCVmixChannels(), Image< T >::getHeight(), Image< T >::getWidth(), inplaceAttenuateBorders(), RASFMT_PNM, Raster::ReadGray(), Raster::ReadRGB(), sformat(), sqrt(), and Raster::WriteGray().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVmixChannels	(	Image< FLOAT > &	img0,
		Image< FLOAT > &	img45,
		Image< FLOAT > &	img90,
		Image< FLOAT > &	img135,
		Image< FLOAT > *	Alpha,
		Image< FLOAT > *	Beta,
		Image< FLOAT > *	Gamma
	)			`[inline]`

When called, this mixes the four channels together into a single channle.

This works by finding the normalized difference between the output of two orthogonal channles. Then it averages the results over all the channels specified. This works to find the rotation invariant properties of orientation responses. This is like the other mixChannels however, this works off of Image class images.

Parameters:

	img0	An image filtered at 0 degrees
	img45	An image filtered at 45 degrees
	img90	An image filtered at 90 degrees
	img135	An image filtered at 135 degrees
	Alpha	This is an image of lineyness in an image
	Beta	This is an image of the crossyness in an image
	Gamma	This is the junction results with crisp crosses
	norm1	This is a normalizer in (img0+img90)/norm
	norm2	This is a normalizer in (img45+img135)/norm

Definition at line 780 of file featureClusterVision.C.

References abs(), and getMinMax().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVmixChannels	(	std::vector< std::vector< FLOAT > > *	data,
		int	ch1a,
		int	ch1b,
		int	ch2a,
		int	ch2b,
		typename std::vector< FLOAT > *	outAlpha,
		typename std::vector< FLOAT > *	outBeta,
		typename std::vector< FLOAT > *	outGamma,
		FLOAT	norm1,
		FLOAT	norm2,
		int	size = `0`
	)			`[inline]`

When called, this mixes the four channels together into a single channle.

This works by finding the normalized difference between the output of two orthogonal channles. Then it averages the results over all the channels specified. This works to find the rotation invariant properties of orientation responses.

Parameters:

	data	This is the raw channle input
	ch1a	this is the first channle input
	ch1b	this is the orthogonal channel to ch1a
	ch2a	this is the second channel set to work in
	ch2b	this is the orthogonal channel to ch2a
	outAlpha	this is a vector of the mixed channels strength
	outBeta	this is a vector of the mixed channels line clearity
	outGamma	this is a vector of the mixed channels junction clearity
	norm1	this is a normalizer over ch1a and ch1b (e.g. Max value)
	norm2	this is a normalizer over ch2a and ch2b (e.g. Max value)

Definition at line 728 of file featureClusterVision.C.

References ASSERT.

Referenced by featureClusterVision< FLOAT >::fCVfindMixedChannels(), and featureClusterVision< FLOAT >::fCVgetImageComplexStats().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVrunStandAloneMSBatchFilter ( std::string filename ) [inline]

check stand alone feature extraction

Definition at line 1287 of file featureClusterVision.C.

References Timer::get(), H1, H2, multiScaleBatchFilter(), multiScaleJunctionFilter(), and Timer::reset().

Referenced by featureClusterVision< FLOAT >::fCVstandAloneFeatureTest().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVrunStandAloneMSBatchTest ( std::string filename ) [inline]

check stand alone feature extraction

Definition at line 1424 of file featureClusterVision.C.

References Image< T >::beginw(), FLOAT_NORM_0_255, Timer::get(), Image< T >::getHeight(), Image< T >::getVal(), Image< T >::getWidth(), H1, H2, logsig(), matrixMult(), multiScaleBatchFilter(), multiScaleJunctionFilter(), Timer::reset(), Image< T >::resize(), Image< T >::setVal(), sformat(), and Raster::VisuFloat().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVsaccadeTest	(	std::string	_maskFile,
		std::string	_outFile,
		std::string	_label,
		std::string	_fileName
	)			`[inline]`

A test method to use with raw saccade data.

Definition at line 3142 of file featureClusterVision.C.

References featureClusterVision< FLOAT >::fCVfindMixedChannels(), Timer::get(), and Timer::reset().

template<class FLOAT >

void featureClusterVision< FLOAT >::fCVsetUpfCV	(	OptionManager &	mgr,
		const std::string &	descrName,
		const std::string &	tagName,
		nub::soft_ref< InputFrameSeries > &	_ifs,
		const std::string &	extraArg0
	)			`[inline]`