Extended Pairwise Layered graph class. More...

#include <GraphLayeredExt.h>

Inheritance diagram for DirectGraphicalModels::CGraphLayeredExt:

Collaboration diagram for DirectGraphicalModels::CGraphLayeredExt:

Public Member Functions
	CGraphLayeredExt (IGraphPairwise &graph, word nLayers, byte gType=GRAPH_EDGES_GRID)
	Constructor. More...

virtual	~CGraphLayeredExt (void)=default

void	buildGraph (Size graphSize) override
	Builds a 2D graph of size corresponding to the image resolution. More...

void	setGraph (const Mat &pots) override
	Fills an existing 2D graph with potentials or builds a new 2D graph of size corresponding to pots.size() with potentials. More...

void	addDefaultEdgesModel (float val, float weight=1.0f) override
	Adds default data-independet edge model. More...

void	addDefaultEdgesModel (const Mat &featureVectors, float val, float weight=1.0f) override
	Adds default contrast-sensitive edge model. More...

void	addDefaultEdgesModel (const vec_mat_t &featureVectors, float val, float weight=1.0f) override
	Adds default contrast-sensitive edge model. More...

Size	getSize () const override
	Returns the size of the graph. More...

void	setGraph (const Mat &potBase, const Mat &potOccl)
	Fills the graph nodes with potentials. More...

void	addFeatureVecs (CTrainEdge &edgeTrainer, const Mat &featureVectors, const Mat &gt)
	Adds a block of new feature vectors. More...

void	addFeatureVecs (CTrainEdge &edgeTrainer, const vec_mat_t &featureVectors, const Mat &gt)
	Adds a block of new feature vectors. More...

void	fillEdges (const CTrainEdge &edgeTrainer, const CTrainLink *linkTrainer, const Mat &featureVectors, const vec_float_t &vParams, float edgeWeight=1.0f, float linkWeight=1.0f)
	Fills the graph edges with potentials. More...

void	fillEdges (const CTrainEdge &edgeTrainer, const CTrainLink *linkTrainer, const vec_mat_t &featureVectors, const vec_float_t &vParams, float edgeWeight=1.0f, float linkWeight=1.0f)
	Fills the graph edges with potentials. More...

void	defineEdgeGroup (float A, float B, float C, byte group)
	Assign the edges, which cross the given line to the grop group. More...

void	setEdges (std::optional< byte > group, const Mat &pot)
	Sets potential pot to all edges in the group group. More...

byte	getType (void) const
	Returns the type of the graph. More...

IGraphPairwise &	getGraph (void) const
	Returns the reference to the graph. More...

Public Member Functions inherited from DirectGraphicalModels::CGraphExt
	CGraphExt (void)=default

	CGraphExt (const CGraphExt &)=delete

virtual	~CGraphExt (void)=default

const CGraphExt &	operator= (const CGraphExt &)=delete

Private Attributes
IGraphPairwise &	m_graph
	The graph. More...

const word	m_nLayers
	Number of layers. More...

const byte	m_gType
	Graph type (Ref. graphEdgesType) More...

Size	m_size
	Size of the graph. More...

Detailed Description

Extended Pairwise Layered graph class.

This graph class provides additional functionality, when the multi-layer graph is used for 2D image classification. The implementation is based on Labeling of Partially Occluded Regions via the Multi-Layer CRF paper.

Author: Dr. Sergey Kosov, serge.nosp@m.y.ko.nosp@m.sov@p.nosp@m.roje.nosp@m.ct-10.nosp@m..de

Definition at line 33 of file GraphLayeredExt.h.

Constructor & Destructor Documentation

◆ CGraphLayeredExt()

DirectGraphicalModels::CGraphLayeredExt::CGraphLayeredExt	(	IGraphPairwise &	graph,
		word	nLayers,
		byte	gType = `GRAPH_EDGES_GRID`
	)

inline

Constructor.

Parameters

graph	The graph
nLayers	The number of layers
gType	The graph type. (Ref. graphEdgesType)

Definition at line 42 of file GraphLayeredExt.h.

◆ ~CGraphLayeredExt()

virtual DirectGraphicalModels::CGraphLayeredExt::~CGraphLayeredExt ( void )

virtualdefault

Member Function Documentation

◆ addDefaultEdgesModel() [1/3]

void DirectGraphicalModels::CGraphLayeredExt::addDefaultEdgesModel	(	float	val,
		float	weight = `1.0f`
	)

overridevirtual

Adds default data-independet edge model.

Parameters

val	Value, specifying the smoothness strength
weight	The weighting parameter

Implements DirectGraphicalModels::CGraphExt.

Definition at line 67 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

◆ addDefaultEdgesModel() [2/3]

void DirectGraphicalModels::CGraphLayeredExt::addDefaultEdgesModel	(	const Mat &	featureVectors,
		float	val,
		float	weight = `1.0f`
	)

overridevirtual

Adds default contrast-sensitive edge model.

Parameters

featureVectors	Multi-channel matrix, each element of which is a multi-dimensinal point: Mat(type: CV_8UC<nFeatures>)
val	Value, specifying the smoothness strength
weight	The weighting parameter

Implements DirectGraphicalModels::CGraphExt.

Definition at line 75 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

◆ addDefaultEdgesModel() [3/3]

void DirectGraphicalModels::CGraphLayeredExt::addDefaultEdgesModel	(	const vec_mat_t &	featureVectors,
		float	val,
		float	weight = `1.0f`
	)

overridevirtual

Adds default contrast-sensitive edge model.

Parameters

featureVectors	Vector of size nFeatures, each element of which is a single feature - image: Mat(type: CV_8UC1)
val	Value, specifying the smoothness strength
weight	The weighting parameter

Implements DirectGraphicalModels::CGraphExt.

Definition at line 84 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

◆ addFeatureVecs() [1/2]

void DirectGraphicalModels::CGraphLayeredExt::addFeatureVecs	(	CTrainEdge &	edgeTrainer,
		const Mat &	featureVectors,
		const Mat &	gt
	)

Adds a block of new feature vectors.

This function may be used only for basic graphical models, built with the CGraphExt::build() method. It extracts pairs of feature vectors with corresponding ground-truth values from blocks featureVectors and gt, according to the graph structure, provided during the class construction via gType argument

Parameters

edgeTrainer	A pointer to the edge trainer
featureVectors	Multi-channel matrix, each element of which is a multi-dimensinal point: Mat(type: CV_8UC<nFeatures>)
gt	Matrix, each element of which is a ground-truth state (class)

Definition at line 151 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

◆ addFeatureVecs() [2/2]

void DirectGraphicalModels::CGraphLayeredExt::addFeatureVecs	(	CTrainEdge &	edgeTrainer,
		const vec_mat_t &	featureVectors,
		const Mat &	gt
	)

Adds a block of new feature vectors.

This function may be used only for basic graphical models, built with the CGraphExt::build() method. It extracts pairs of feature vectors with corresponding ground-truth values from blocks featureVectors and gt, according to the graph structure, provided during the class construction via gType argument

Parameters

edgeTrainer	A pointer to the edge trainer
featureVectors	Vector of size nFeatures, each element of which is a single feature - image: Mat(type: CV_8UC1)
gt	Matrix, each element of which is a ground-truth state (class)

Definition at line 200 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

◆ buildGraph()

void DirectGraphicalModels::CGraphLayeredExt::buildGraph ( Size graphSize )

overridevirtual

Builds a 2D graph of size corresponding to the image resolution.

All edges in graph will have group id 0 except the edges connecting different layers (links), which will have group id 1. When called multiple times, previouse graph structure is always replaced.

Parameters

graphSize The size of the graph (image resolution)

Implements DirectGraphicalModels::CGraphExt.

Definition at line 13 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ defineEdgeGroup()

void DirectGraphicalModels::CGraphLayeredExt::defineEdgeGroup	(	float	A,
		float	B,
		float	C,
		byte	group
	)

Assign the edges, which cross the given line to the grop group.

The line is given by the equation: Ax + By + C = 0. A and B are not both equal to zero.

Parameters

A	Constant line parameter
B	Constant line parameter
C	Constant line parameter
group	New group ID

Definition at line 410 of file GraphLayeredExt.cpp.

◆ fillEdges() [1/2]

void DirectGraphicalModels::CGraphLayeredExt::fillEdges	(	const CTrainEdge &	edgeTrainer,
		const CTrainLink *	linkTrainer,
		const Mat &	featureVectors,
		const vec_float_t &	vParams,
		float	edgeWeight = `1.0f`,
		float	linkWeight = `1.0f`
	)

Fills the graph edges with potentials.

This function uses edgeTrainer class in oerder to achieve edge potentials from feature vectors, stored in featureVectors and fills with them the graph edges

This function supports PPL

Parameters

edgeTrainer	A pointer to the edge trainer
linkTrainer	A pointer to tht link (inter-layer edge) trainer
featureVectors	Multi-channel matrix, each element of which is a multi-dimensinal point: Mat(type: CV_8UC<nFeatures>)
vParams	Array of control parameters. Please refer to the concrete model implementation of the CTrainEdge::calculateEdgePotentials() function for more details
edgeWeight	The weighting parameter for (within-layer) edges
linkWeight	The weighting parameter for (inter-layer) edges, i.e. links

Definition at line 253 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ fillEdges() [2/2]

void DirectGraphicalModels::CGraphLayeredExt::fillEdges	(	const CTrainEdge &	edgeTrainer,
		const CTrainLink *	linkTrainer,
		const vec_mat_t &	featureVectors,
		const vec_float_t &	vParams,
		float	edgeWeight = `1.0f`,
		float	linkWeight = `1.0f`
	)

Fills the graph edges with potentials.

This function uses edgeTrainer class in oerder to achieve edge potentials from feature vectors, stored in featureVectors and fills with them the graph edges

This function supports PPL

Parameters

edgeTrainer	A pointer to the edge trainer
linkTrainer	A pointer to tht link (inter-layer edge) trainer
featureVectors	Vector of size nFeatures, each element of which is a single feature - image: Mat(type: CV_8UC1)
vParams	Array of control parameters. Please refer to the concrete model implementation of the CTrainEdge::calculateEdgePotentials() function for more details
edgeWeight	The weighting parameter for (within-layer) edges
linkWeight	The weighting parameter for (inter-layer) edges, i.e. links

Definition at line 328 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

◆ getGraph()

IGraphPairwise& DirectGraphicalModels::CGraphLayeredExt::getGraph ( void ) const

inline

Returns the reference to the graph.

Returns: The reference to the graph

Definition at line 158 of file GraphLayeredExt.h.

◆ getSize()

Size DirectGraphicalModels::CGraphLayeredExt::getSize ( ) const

inlineoverridevirtual

Returns the size of the graph.

Returns: The size of the Graph

Implements DirectGraphicalModels::CGraphExt.

Definition at line 74 of file GraphLayeredExt.h.

◆ getType()

byte DirectGraphicalModels::CGraphLayeredExt::getType ( void ) const

inline

Returns the type of the graph.

Returns: The type of the graph (Ref. graphEdgesType)

Definition at line 153 of file GraphLayeredExt.h.

◆ setEdges()

void DirectGraphicalModels::CGraphLayeredExt::setEdges	(	std::optional< byte >	group,
		const Mat &	pot
	)

Sets potential pot to all edges in the group group.

Parameters

pot	Edge potential matrix: Mat(size: nStates x nStates; type: CV_32FC1)
group	The edge group ID

Definition at line 460 of file GraphLayeredExt.cpp.

◆ setGraph() [1/2]

void DirectGraphicalModels::CGraphLayeredExt::setGraph ( const Mat & pots )

overridevirtual

Fills an existing 2D graph with potentials or builds a new 2D graph of size corresponding to pots.size() with potentials.

If the graph was not build beforehand, or the size of existing graph does not correspond to pots.size() this function calls first

buildGraph(pots.size())

Parameters

pots	A block of node potentials: Mat(size: image width x image height; type: CV_32FC(nStates)). It may be obtained by: CTrainNode::getNodePotentials()

Implements DirectGraphicalModels::CGraphExt.

Definition at line 61 of file GraphLayeredExt.cpp.

◆ setGraph() [2/2]

void DirectGraphicalModels::CGraphLayeredExt::setGraph	(	const Mat &	potBase,
		const Mat &	potOccl
	)

Fills the graph nodes with potentials.

If the graph was not build beforehand, this function calls first

buildGraph(potBase.size())

This function supports PPL

Parameters

potBase	A block of potentials for the base layer: Mat(type: CV_32FC(nStatesBase))
potOccl	A block of potentials for the occlusion layer: Mat(type: CV_32FC(nStatesOccl))

Definition at line 93 of file GraphLayeredExt.cpp.

Here is the call graph for this function:

Member Data Documentation

◆ m_graph

IGraphPairwise& DirectGraphicalModels::CGraphLayeredExt::m_graph

private

The graph.

Definition at line 162 of file GraphLayeredExt.h.

◆ m_gType

const byte DirectGraphicalModels::CGraphLayeredExt::m_gType

private

Graph type (Ref. graphEdgesType)

Definition at line 164 of file GraphLayeredExt.h.

◆ m_nLayers

const word DirectGraphicalModels::CGraphLayeredExt::m_nLayers

private

Number of layers.

Definition at line 163 of file GraphLayeredExt.h.

◆ m_size

Size DirectGraphicalModels::CGraphLayeredExt::m_size

private

Size of the graph.

Definition at line 165 of file GraphLayeredExt.h.

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

D:/Projects/DGM/modules/DGM/GraphLayeredExt.h
D:/Projects/DGM/modules/DGM/GraphLayeredExt.cpp

Public Member Functions

Private Attributes

Detailed Description

Constructor & Destructor Documentation

◆ CGraphLayeredExt()

◆ ~CGraphLayeredExt()

Member Function Documentation

◆ addDefaultEdgesModel() [1/3]

◆ addDefaultEdgesModel() [2/3]

◆ addDefaultEdgesModel() [3/3]

◆ addFeatureVecs() [1/2]

◆ addFeatureVecs() [2/2]

◆ buildGraph()

◆ defineEdgeGroup()

◆ fillEdges() [1/2]

◆ fillEdges() [2/2]

◆ getGraph()

◆ getSize()

◆ getType()

◆ setEdges()

◆ setGraph() [1/2]

◆ setGraph() [2/2]

Member Data Documentation

◆ m_graph

◆ m_gType

◆ m_nLayers

◆ m_size