Hierarchical kmeans index. More...

#include <opencv2/flann/kmeans_index.h>

Inheritance diagram for cvflann::KMeansIndex< Distance >:

Collaboration diagram for cvflann::KMeansIndex< Distance >:

Classes
class	KMeansDistanceComputer

Public Types
typedef void(KMeansIndex::*	centersAlgFunction) (int, int , int, int , int &)

typedef Distance::ResultType	DistanceType

typedef Distance::ElementType	ElementType

Public Member Functions
	KMeansIndex (const Matrix< ElementType > &inputData, const IndexParams &params=KMeansIndexParams(), Distance d=Distance())
	Index constructor. More...

	KMeansIndex (const KMeansIndex &)

virtual	~KMeansIndex ()
	Index destructor. More...

void	buildIndex () CV_OVERRIDE
	Builds the index. More...

void	chooseCentersGonzales (int k, int indices, int indices_length, int centers, int &centers_length)
	Chooses the initial centers in the k-means using Gonzales' algorithm so that the centers are spaced apart from each other. More...

void	chooseCentersKMeanspp (int k, int indices, int indices_length, int centers, int &centers_length)
	Chooses the initial centers in the k-means using the algorithm proposed in the KMeans++ paper: Arthur, David; Vassilvitskii, Sergei - k-means++: The Advantages of Careful Seeding. More...

void	chooseCentersRandom (int k, int indices, int indices_length, int centers, int &centers_length)
	Chooses the initial centers in the k-means clustering in a random manner. More...

void	findNeighbors (ResultSet< DistanceType > &result, const ElementType *vec, const SearchParams &searchParams) CV_OVERRIDE
	Find set of nearest neighbors to vec. More...

int	getClusterCenters (Matrix< DistanceType > &centers)
	Clustering function that takes a cut in the hierarchical k-means tree and return the clusters centers of that clustering. More...

IndexParams	getParameters () const CV_OVERRIDE

flann_algorithm_t	getType () const CV_OVERRIDE

virtual void	knnSearch (const Matrix< ElementType > &queries, Matrix< int > &indices, Matrix< DistanceType > &dists, int knn, const SearchParams &params)
	Perform k-nearest neighbor search. More...

void	loadIndex (FILE *stream) CV_OVERRIDE
	Loads the index from a stream. More...

KMeansIndex &	operator= (const KMeansIndex &)

virtual int	radiusSearch (const Matrix< ElementType > &query, Matrix< int > &indices, Matrix< DistanceType > &dists, float radius, const SearchParams &params)
	Perform radius search. More...

void	saveIndex (FILE *stream) CV_OVERRIDE
	Saves the index to a stream. More...

void	set_cb_index (float index)

size_t	size () const CV_OVERRIDE
	Returns size of index. More...

int	usedMemory () const CV_OVERRIDE
	Computes the inde memory usage Returns: memory used by the index. More...

size_t	veclen () const CV_OVERRIDE
	Returns the length of an index feature. More...

Public Attributes
centersAlgFunction	chooseCenters
	The function used for choosing the cluster centers. More...

Detailed Description

template<typename Distance>
class cvflann::KMeansIndex< Distance >

Hierarchical kmeans index.

Contains a tree constructed through a hierarchical kmeans clustering and other information for indexing a set of points for nearest-neighbour matching.

Member Typedef Documentation

◆ centersAlgFunction

template<typename Distance>

typedef void(KMeansIndex::* cvflann::KMeansIndex< Distance >::centersAlgFunction) (int, int *, int, int *, int &)

◆ DistanceType

template<typename Distance>

typedef Distance::ResultType cvflann::KMeansIndex< Distance >::DistanceType

◆ ElementType

template<typename Distance>

typedef Distance::ElementType cvflann::KMeansIndex< Distance >::ElementType

Constructor & Destructor Documentation

◆ KMeansIndex() [1/2]

template<typename Distance>

cvflann::KMeansIndex< Distance >::KMeansIndex	(	const Matrix< ElementType > &	inputData,
		const IndexParams &	params = `KMeansIndexParams()`,
		Distance	d = `Distance()`
	)

inline

Index constructor.

Params: inputData = dataset with the input features params = parameters passed to the hierarchical k-means algorithm

References cvflann::KMeansIndex< Distance >::chooseCentersGonzales(), cvflann::KMeansIndex< Distance >::chooseCentersKMeanspp(), cvflann::KMeansIndex< Distance >::chooseCentersRandom(), cvflann::FLANN_CENTERS_GONZALES, cvflann::FLANN_CENTERS_KMEANSPP, cvflann::FLANN_CENTERS_RANDOM, cvflann::get_param(), and cv::max().

Here is the call graph for this function:

◆ KMeansIndex() [2/2]

template<typename Distance>

cvflann::KMeansIndex< Distance >::KMeansIndex ( const KMeansIndex< Distance > & )

◆ ~KMeansIndex()

template<typename Distance>

virtual cvflann::KMeansIndex< Distance >::~KMeansIndex ( )

inlinevirtual

Index destructor.

Release the memory used by the index.

Member Function Documentation

◆ buildIndex()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::buildIndex ( )

inlinevirtual

Builds the index.

Implements cvflann::NNIndex< Distance >.

Referenced by cvflann::AutotunedIndex< Distance >::getType(), and cvflann::hierarchicalClustering().

Here is the caller graph for this function:

◆ chooseCentersGonzales()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::chooseCentersGonzales	(	int	k,
		int *	indices,
		int	indices_length,
		int *	centers,
		int &	centers_length
	)

inline

Chooses the initial centers in the k-means using Gonzales' algorithm so that the centers are spaced apart from each other.

Params: k = number of centers vecs = the dataset of points indices = indices in the dataset Returns:

References cvflann::rand_int().

Referenced by cvflann::KMeansIndex< Distance >::KMeansIndex().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ chooseCentersKMeanspp()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::chooseCentersKMeanspp	(	int	k,
		int *	indices,
		int	indices_length,
		int *	centers,
		int &	centers_length
	)

inline

Chooses the initial centers in the k-means using the algorithm proposed in the KMeans++ paper: Arthur, David; Vassilvitskii, Sergei - k-means++: The Advantages of Careful Seeding.

Implementation of this function was converted from the one provided in Arthur's code.

Params: k = number of centers vecs = the dataset of points indices = indices in the dataset Returns:

References cv::min(), cvflann::rand_double(), and cvflann::rand_int().

Referenced by cvflann::KMeansIndex< Distance >::KMeansIndex().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ chooseCentersRandom()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::chooseCentersRandom	(	int	k,
		int *	indices,
		int	indices_length,
		int *	centers,
		int &	centers_length
	)

inline

Chooses the initial centers in the k-means clustering in a random manner.

Params: k = number of centers vecs = the dataset of points indices = indices in the dataset indices_length = length of indices vector

References cvflann::UniqueRandom::next().

Referenced by cvflann::KMeansIndex< Distance >::KMeansIndex().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ findNeighbors()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::findNeighbors	(	ResultSet< DistanceType > &	result,
		const ElementType *	vec,
		const SearchParams &	searchParams
	)

inlinevirtual

Find set of nearest neighbors to vec.

Their indices are stored inside the result object.

Params: result = the result object in which the indices of the nearest-neighbors are stored vec = the vector for which to search the nearest neighbors searchParams = parameters that influence the search algorithm (checks, cb_index)

Implements cvflann::NNIndex< Distance >.

References cvflann::FLANN_CHECKS_UNLIMITED, cvflann::get_param(), cvflann::BranchStruct< T, DistanceType >::node, and cvflann::Heap< T >::popMin().

Here is the call graph for this function:

◆ getClusterCenters()

template<typename Distance>

int cvflann::KMeansIndex< Distance >::getClusterCenters ( Matrix< DistanceType > & centers )

inline

Clustering function that takes a cut in the hierarchical k-means tree and return the clusters centers of that clustering.

Params: numClusters = number of clusters to have in the clustering computed Returns: number of cluster centers

References cvflann::Logger::info(), and cvflann::Matrix< T >::rows.

Referenced by cvflann::hierarchicalClustering().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ getParameters()

template<typename Distance>

IndexParams cvflann::KMeansIndex< Distance >::getParameters ( ) const

inlinevirtual

Returns: The index parameters

Implements cvflann::NNIndex< Distance >.

References cvflann::ResultSet< DistanceType >::addPoint(), cv::AutoBuffer< _Tp, fixed_size >::data(), cvflann::ResultSet< DistanceType >::full(), cvflann::Heap< T >::insert(), cvflann::load_value(), cv::max(), cv::mean(), cv::parallel_for_(), cvflann::save_value(), cv::sort(), cv::sqrt(), cv::sum(), cv::swap(), and cvflann::ResultSet< DistanceType >::worstDist().

Here is the call graph for this function:

◆ getType()

template<typename Distance>

flann_algorithm_t cvflann::KMeansIndex< Distance >::getType ( ) const

inlinevirtual

Returns: The index type (kdtree, kmeans,...)

Implements cvflann::NNIndex< Distance >.

References cvflann::FLANN_INDEX_KMEANS.

◆ knnSearch()

template<typename Distance>

virtual void cvflann::NNIndex< Distance >::knnSearch	(	const Matrix< ElementType > &	queries,
		Matrix< int > &	indices,
		Matrix< DistanceType > &	dists,
		int	knn,
		const SearchParams &	params
	)

inlinevirtualinherited

Perform k-nearest neighbor search.

Parameters

[in]	queries	The query points for which to find the nearest neighbors
[out]	indices	The indices of the nearest neighbors found
[out]	dists	Distances to the nearest neighbors found
[in]	knn	Number of nearest neighbors to return
[in]	params	Search parameters

Reimplemented in cvflann::Index< Distance >, cvflann::Index< cv::L2< ElementType > >, cvflann::Index< cv::L1< ElementType > >, cvflann::KDTreeSingleIndex< Distance >, and cvflann::LshIndex< Distance >.

◆ loadIndex()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::loadIndex ( FILE * stream )

inlinevirtual

Loads the index from a stream.

Parameters

stream The stream from which the index is loaded

Implements cvflann::NNIndex< Distance >.

References cvflann::load_value().

Here is the call graph for this function:

◆ operator=()

template<typename Distance>

KMeansIndex& cvflann::KMeansIndex< Distance >::operator= ( const KMeansIndex< Distance > & )

◆ radiusSearch()

template<typename Distance>

virtual int cvflann::NNIndex< Distance >::radiusSearch	(	const Matrix< ElementType > &	query,
		Matrix< int > &	indices,
		Matrix< DistanceType > &	dists,
		float	radius,
		const SearchParams &	params
	)

inlinevirtualinherited

Perform radius search.

Parameters

[in]	query	The query point
[out]	indices	The indinces of the neighbors found within the given radius
[out]	dists	The distances to the nearest neighbors found
[in]	radius	The radius used for search
[in]	params	Search parameters

Returns: Number of neighbors found

Reimplemented in cvflann::Index< Distance >, cvflann::Index< cv::L2< ElementType > >, and cvflann::Index< cv::L1< ElementType > >.

◆ saveIndex()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::saveIndex ( FILE * stream )

inlinevirtual

Saves the index to a stream.

Parameters

stream The stream to save the index to

Implements cvflann::NNIndex< Distance >.

References cvflann::save_value().

Here is the call graph for this function:

◆ set_cb_index()

template<typename Distance>

void cvflann::KMeansIndex< Distance >::set_cb_index ( float index )

inline

Referenced by cvflann::AutotunedIndex< Distance >::getType().

Here is the caller graph for this function:

◆ size()

template<typename Distance>

size_t cvflann::KMeansIndex< Distance >::size ( ) const

inlinevirtual

Returns size of index.

Implements cvflann::NNIndex< Distance >.

◆ usedMemory()

template<typename Distance>

int cvflann::KMeansIndex< Distance >::usedMemory ( ) const

inlinevirtual

Computes the inde memory usage Returns: memory used by the index.

Implements cvflann::NNIndex< Distance >.

Referenced by cvflann::AutotunedIndex< Distance >::getType().

Here is the caller graph for this function:

◆ veclen()

template<typename Distance>

size_t cvflann::KMeansIndex< Distance >::veclen ( ) const

inlinevirtual

Returns the length of an index feature.

Implements cvflann::NNIndex< Distance >.

Member Data Documentation

◆ chooseCenters

template<typename Distance>

centersAlgFunction cvflann::KMeansIndex< Distance >::chooseCenters

The function used for choosing the cluster centers.

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

opencv2/flann/kmeans_index.h

Classes

Public Types

Public Member Functions

Public Attributes

Detailed Description

template<typename Distance> class cvflann::KMeansIndex< Distance >

Member Typedef Documentation

◆ centersAlgFunction

◆ DistanceType

◆ ElementType

Constructor & Destructor Documentation

◆ KMeansIndex() [1/2]

◆ KMeansIndex() [2/2]

◆ ~KMeansIndex()

Member Function Documentation

◆ buildIndex()

◆ chooseCentersGonzales()

◆ chooseCentersKMeanspp()

◆ chooseCentersRandom()

◆ findNeighbors()

◆ getClusterCenters()

◆ getParameters()

◆ getType()

◆ knnSearch()

◆ loadIndex()

◆ operator=()

◆ radiusSearch()

◆ saveIndex()

◆ set_cb_index()

◆ size()

◆ usedMemory()

◆ veclen()

Member Data Documentation

◆ chooseCenters

template<typename Distance>
class cvflann::KMeansIndex< Distance >