Classification

Typedefs
typedef double	TTransformation[16]
	Generic description of a transformation. More...

Functions
cvbbool_t	PMClassify (TCLF simpleClf, IMG img, cvbdim_t X, cvbdim_t Y, char *szClassLabel, size_t len, double &confidence, double *confidences)
	Classifies the neighborhood of a point (X,Y) in an image according to the specified classifier. More...
cvbbool_t	PMClassifyW (TCLF simpleClf, IMG img, cvbdim_t X, cvbdim_t Y, wchar_t *szClassLabel, size_t len, double &confidence, double *confidences)
	Classifies the neighborhood of a point (X,Y) in an image according to the specified classifier. More...
void	PMFormatTransformation (TSCLF searchClf, double(&t)[16], TGeometrySpec &specification)
	Translate a geometry description. More...
cvbbool_t	PMGetClfClassLabel (TCLF clf, cvbval_t classIndex, char *szClassLabel)
	Get the class label corresponding to a given index. More...
cvbbool_t	PMGetClfClassLabelW (TCLF clf, cvbval_t classIndex, wchar_t *szClassLabel)
	Get the class label corresponding to a given index. More...
void	PMGetClfTrainParams (TCLF clf, TTrainParams &params)
	Returns the training parameters that have been used during classifier creation in a TTrainParams structure. More...
cvbval_t	PMGetGranularity (const char *szCode)
	The function returns the granularity (see tutorial) associated with a preprocessing code. More...
cvbval_t	PMGetNumClasses (TCLF clf)
	Returns the number of classes trained in a classifier if its usage is CU_ClassifyOneVersusAll or CU_ClassifyOneVersusOne. More...
cvbval_t	PMGetOutputDimension (TCLF clf)
	Returns the number of parameters estimated for numerical prediction by the classifier if the classifier's usage mode is CU_Regression. More...
cvbbool_t	PMInterpretSpecification (TSCLF searchClf, const TGeometrySpec &specification, double(&t)[16])
	Translate a geometry description. More...
cvbbool_t	PMPredictVector (TCLF simpleClf, IMG img, cvbdim_t X, cvbdim_t Y, double *resultVector)
	Numerical prediction of vector-valued function depending on the neighborhood of a point (X,Y) in an image according to the specified classifier. More...

Detailed Description

Typedef Documentation

TTransformation

typedef double TTransformation[16]

Generic description of a transformation.

The TTransformation type generically describes transformations as an array of double values. The interpretation of these double values is dependent on the actually described transformation, and therefore the type TTransformation is usually never used directly. It is instead defined by the results of PMGridSearch or PMFormatTransformation and used in PMSearchResultToImage as well as PMInterpretSpecification.

Function Documentation

PMClassify()

cvbbool_t PMClassify	(	TCLF	simpleClf,
		IMG	img,
		cvbdim_t	X,
		cvbdim_t	Y,
		char *	szClassLabel,
		size_t	len,
		double &	confidence,
		double *	confidences
	)

Classifies the neighborhood of a point (X,Y) in an image according to the specified classifier.

The classifier usage setting (retrieve with PMGetClfTrainParams) must be either CU_ClassifyOneVersusAll or CU_ClassifyOneVersusOne.

Parameters

[in]	simpleClf	Pointer to classifier.
[in]	img	Image handle.
[in]	X	X coordinate of the point in the image to classify.
[in]	Y	Y coordinate of the point in the image to classify.
[in]	szClassLabel	Pointer to a string buffer where the resulting class label is returned. Generally, class labels are limited to up to 255 characters (not including the terminating zero).
[in]	len	Length of the class label string buffer pointed to by szClassLabel.
[out]	confidence	Maximal confidence.
[in]	confidences	Array where the confidence may be returned for each class label. Set to NULL if the individual confidences should not be returned; otherwise have confidences point to an array of PMGetNumClasses * sizeof(double) bytes.

Returns

TRUE if classification was successful, FALSE if not

PMClassifyW()

cvbbool_t PMClassifyW	(	TCLF	simpleClf,
		IMG	img,
		cvbdim_t	X,
		cvbdim_t	Y,
		wchar_t *	szClassLabel,
		size_t	len,
		double &	confidence,
		double *	confidences
	)

Classifies the neighborhood of a point (X,Y) in an image according to the specified classifier.

The classifier usage setting (retrieve with PMGetClfTrainParams) must be either CU_ClassifyOneVersusAll or CU_ClassifyOneVersusOne.

Parameters

[in]	simpleClf	Pointer to classifier.
[in]	img	Image handle.
[in]	X	X coordinate of the point in the image to classify.
[in]	Y	Y coordinate of the point in the image to classify.
[in]	szClassLabel	Pointer to a string buffer where the resulting class label is returned. Generally, class labels are limited to up to 255 characters (not including the terminating zero).
[in]	len	Length of the class label string buffer pointed to by szClassLabel.
[out]	confidence	Maximal confidence.
[in]	confidences	Array where the confidence may be returned for each class label. Set to NULL if the individual confidences should not be returned; otherwise have confidences point to an array of PMGetNumClasses * sizeof(double) bytes.

Returns

TRUE if classification was successful, FALSE if not

PMFormatTransformation()

void PMFormatTransformation	(	TSCLF	searchClf,
		double(&)	t[16],
		TGeometrySpec &	specification
	)

Translate a geometry description.

This method translates a classifier's geometry description as given by a TTransformation array into a readable geometry description as given by a TGeometrySpec struct, taking into account the classifier's view on geometries.

Parameters

[in]	searchClf	The search classifiers whose geometry interpretation is to be used for formatting.
[in]	t	Classifier-dependent geometry to be interpreted.
[out]	specification	Struct that makes the classifier's view at the parameter t readable

PMGetClfClassLabel()

cvbbool_t PMGetClfClassLabel	(	TCLF	clf,
		cvbval_t	classIndex,
		char *	szClassLabel
	)

Get the class label corresponding to a given index.

Attention

The buffer pointed to by szClassLabel must be big enough to hold the entire class label, otherwise a write access violation may occur. To be on the safe side, alway allocate at least 256 bytes.

Parameters

[in]	clf	Handle of the classifier to be queried.
[in]	classIndex	Index of the class in question. Must be nonnegative and smaller than the return value of PMGetNumClasses.
[in]	szClassLabel	Pointer to a buffer that will receive the class label string (zero-terminated!).

Returns

TRUE if the call was successful (i.e. all parameters are valid) FALSE otherwise.

PMGetClfClassLabelW()

cvbbool_t PMGetClfClassLabelW	(	TCLF	clf,
		cvbval_t	classIndex,
		wchar_t *	szClassLabel
	)

Get the class label corresponding to a given index.

Attention

The buffer pointed to by szClassLabel must be big enough to hold the entire class label, otherwise a write access violation may occur. To be on the safe side, alway allocate at least 256 bytes.

Parameters

[in]	clf	Handle of the classifier to be queried.
[in]	classIndex	Index of the class in question. Must be nonnegative and smaller than the return value of PMGetNumClasses.
[in]	szClassLabel	Pointer to a buffer that will receive the class label string (zero-terminated!).

Returns

TRUE if the call was successful (i.e. all parameters are valid) FALSE otherwise.

PMGetClfTrainParams()

void PMGetClfTrainParams	(	TCLF	clf,
		TTrainParams &	params
	)

Returns the training parameters that have been used during classifier creation in a TTrainParams structure.

Note that this function should not be called on a search classifier. To retrieve the parameters that have been used for a search classifier please use PMGetSearchClfTrainParams.

Parameters

[in]	clf	pointer to classifier.
[out]	params	The training parameters used during generation of the classifier.

PMGetGranularity()

cvbval_t PMGetGranularity

(

const char *

szCode

)

The function returns the granularity (see tutorial) associated with a preprocessing code.

Parameters

[in]

szCode

Pointer to 0-terminated string containing the preprocessing code.

Returns

Divisor for width and height of feature window.

PMGetNumClasses()

cvbval_t PMGetNumClasses

(

TCLF

clf

)

Returns the number of classes trained in a classifier if its usage is CU_ClassifyOneVersusAll or CU_ClassifyOneVersusOne.

Otherwise it returns zero.

Parameters

[in]

clf

Pointer to classifier.

Returns

Number of classes represented in classifier (or zero if the classifier has been trained for regression).

PMGetOutputDimension()

cvbval_t PMGetOutputDimension

(

TCLF

clf

)

Returns the number of parameters estimated for numerical prediction by the classifier if the classifier's usage mode is CU_Regression.

For classifiers generated with CU_ClassifyOneVersusAll the returned value is the number of classes; for classifiers generated with CU_ClassifyOneVersusOne the returned value is PMGetNumClasses*(PMGetNumClasses-1)/2 (the number of possible pairs of classes).

Parameters

[in]

clf

Pointer to classifier.

Returns

Number of parameters estimated for numerical prediction.

PMInterpretSpecification()

cvbbool_t PMInterpretSpecification	(	TSCLF	searchClf,
		const TGeometrySpec &	specification,
		double(&)	t[16]
	)

Translate a geometry description.

This method translates a geometry description as given by a TGeometrySpec structure into an array of doubles that corresponds to the specified classifier's view on geometries. This is necessary for example when using the function PMInspect or PMInspectD to correctly specify the starting point for the iteration.

Parameters

[in]	searchClf	The search classifiers whose geometry interpretation is to be used for interpretation.
[in]	specification	Geometry description to be interpreted.
[out]	t	Array of 16 doubles that describes the classifier's interpretation of the specification parameter

Returns

TRUE if reformatting succeeded, FALSE if interpretation was not possible because the classifier referenced by searchClf is invalid.

PMPredictVector()

cvbbool_t PMPredictVector	(	TCLF	simpleClf,
		IMG	img,
		cvbdim_t	X,
		cvbdim_t	Y,
		double *	resultVector
	)

Numerical prediction of vector-valued function depending on the neighborhood of a point (X,Y) in an image according to the specified classifier.

The classifier usage setting (retrieve with PMGetClfTrainParams) must be CU_Regression for this function to be able to work with the classifier.

Parameters

[in]	simpleClf	Pointer to classifier.
[in]	img	Image handle.
[in]	X	X coordinate of the point in the image to classify.
[in]	Y	Y coordinate of the point in the image to classify.
[out]	resultVector	Pointer to a double array to be filled with the components of the prediction result vector. Length of the array must be PMGetOutputDimension * sizeof(double).

Returns

TRUE if numerical prediction was successful, FALSE if not.