TrainingSet Class Reference

Manage a paired set of eye position data and input feature vectors. More...

#include <TIGS/TrainingSet.H>

Inheritance diagram for TrainingSet:

[legend]

Collaboration diagram for TrainingSet:

[legend]

List of all members.

Classes
struct	PosGroup
Public Member Functions
	TrainingSet (OptionManager &mgr, const std::string &fx_type)
Dims	scaledInputDims () const
size_t	numPositions () const
int	p2p (const int i, const int j) const
int	p2p (const Point2D< int > &p) const
Image< float >	recordSample (const Point2D< int > &loc, const Image< float > &features)
void	load (const std::string &pfx)
	Load feature and position vectors from a pair of .pfm files (such as those produced by save()).
void	loadRebalanced (const std::string &pfx)
	Like load(), but rebalance the training set so that eye positions are included equally often.
void	save (const std::string &pfx)
Image< float >	getFeatures () const
Image< float >	getPositions () const
uint	inputReduction () const
const std::string &	fxType () const

---

Detailed Description

Manage a paired set of eye position data and input feature vectors.

Definition at line 48 of file TrainingSet.H.

---

Member Function Documentation

void TrainingSet::load

(

const std::string &

pfx

)

Load feature and position vectors from a pair of .pfm files (such as those produced by save()).

<pfx>-features.pfm should contain the features associated with a series of frames, with nrows=NFRAMES and ncols=NFEATURES where NFEATURES is the number of features in the feature vector generated by whichever feature extractor produced this training set

<pfx>-positions.pfm should contain the eye position vectors associated with the same series of frames, with nrows=NFRAMES and ncols=W*H where W*H is the number of pixels in each image of the frame sequence

Definition at line 212 of file TrainingSet.C.

References ASSERT, Image< T >::begin(), blurAndDecY(), Image< T >::end(), Image< T >::getHeight(), OModelParam< T >::getVal(), Image< T >::getWidth(), invt_allocation_release_free_mem(), loadRebalanced(), RASFMT_PFM, and Raster::ReadFloat().

void TrainingSet::loadRebalanced

(

const std::string &

pfx

)

Like load(), but rebalance the training set so that eye positions are included equally often.

Definition at line 271 of file TrainingSet.C.

References ASSERT, Image< T >::begin(), Image< T >::end(), Image< T >::getArrayPtr(), Image< T >::getHeight(), Image< T >::getVal(), OModelParam< T >::getVal(), Image< T >::getWidth(), invt_allocation_release_free_mem(), RASFMT_PFM, Raster::ReadFloat(), and ZEROS.

Referenced by load().

---

The documentation for this class was generated from the following files:

• src/TIGS/TrainingSet.H
• src/TIGS/TrainingSet.C