BinaryClassifierOracle< FeatureGenerator, ClassifierPolicy, Metric, Aggregator > Struct Template Reference

#include <BinaryClassifierOracle.h>

Public Types
typedef FeatureGenerator::FeatureType	FeatureType
	The Feature Extracted by FeatureGenerator.
typedef BinaryClassifier < FeatureType, ClassifierPolicy >	WeakClassifierType
	The weak classifier provided by this oracle.
typedef BoostableClassifier < WeakClassifierType >	ClassifierType
	Additive Regression Final Classifier.

Public Member Functions
template<class SourceDataSet >
void	SetTrainingSet (const SourceDataSet &set)
	Set the training set used to recover the threshold.
void	SetFeatureGenerator (FeatureGenerator &f)
	Associate a Feature Generator to Decision Stump Generator.
DataSetHandle< Aggregator > &	GetTrainingSet ()
	return R/W the training set
void	SetPreloadSize (int n)
void	SetFastHeuristic (bool enable, bool reoptimize, int size)
bool	GetHypothesis (WeakClassifierType &bestH)
double	Fast_Optimize (WeakClassifierType &h)
double	Optimize (WeakClassifierType &h)
	Using current metrics try to recompute parameters associated to this feature.

Detailed Description

template<class FeatureGenerator, class ClassifierPolicy, class Metric, class Aggregator>
struct BinaryClassifierOracle< FeatureGenerator, ClassifierPolicy, Metric, Aggregator >

A base class for training a classifier using Integral Images with some kind of Features (HaarFeature). Find the best threshold of a scalar feature on the trainingset using associated weights, try to maximize AdaBoost W+ param or others Metric

Implements metric based on a list of sorted PatternResponse

Parameters

FeatureGenerator

generates Feature of type FeatureType

Note

FeatureType is the type of feature involved in training

PatternType must have a GetWeight() method implemented.

Member Function Documentation

template<class FeatureGenerator , class ClassifierPolicy , class Metric , class Aggregator >

double BinaryClassifierOracle< FeatureGenerator, ClassifierPolicy, Metric, Aggregator >::Fast_Optimize ( WeakClassifierType &  h )

inline

versione veloce per calcolare la miglior response che fa uso di un istogramma/bin/bucker per evitare la sort TODO serve, per WeakClassifierType, il valore minimo e massimo della response

template<class FeatureType , class ClassifierPolicy , class Metric , class Aggregator >

bool BinaryClassifierOracle< FeatureType, ClassifierPolicy, Metric, Aggregator >::GetHypothesis

(

WeakClassifierType &

bestH

)

Return the best h for current data set. This is the real core of this module: test all FeatureType and produce the bestH classifier with the best of them

template<class FeatureGenerator , class ClassifierPolicy , class Metric , class Aggregator >

void BinaryClassifierOracle< FeatureGenerator, ClassifierPolicy, Metric, Aggregator >::SetFastHeuristic ( bool  enable, bool  reoptimize, int  size  )

inline

Enable/Disable the Fast-optimize heuristic

Parameters

enable	enable/disable
size	number of bins

template<class FeatureGenerator , class ClassifierPolicy , class Metric , class Aggregator >

void BinaryClassifierOracle< FeatureGenerator, ClassifierPolicy, Metric, Aggregator >::SetPreloadSize ( int  n )

inline

Set the preload size

---

The documentation for this struct was generated from the following file:

• include/Oracle/BinaryClassifierOracle.h