Stemmer.Cvb.ShapeFinder2 Namespace Reference

Classes
class	Classifier
	ShapeFinder2 classifier object. More...
class	ClassifierFactory
	Object that aggregates the learning parameters and produces a ShapeFinder2 classifier More...
class	Filter
	Filter functions exported and used by the ShapeFinder library. More...
class	NamespaceDoc
	The namespace and assembly Stemmer.Cvb.ShapeFinder2 contains the classes and definitions needed for using the ShapeFinder 2 functionality from SF.dll. More...
struct	SearchResult
	A single ShapeFinder2 search result. More...

Enumerations
enum	ContrastMode { Normal = 0 , Inverse = 1 }
	Contrast mode for feature extraction. More...
enum	PrecisionMode { NoCorrelation = 0 , CorrelationCoarse = 1 , CorrelationFine = 2 }
	Search mode to be used. More...
enum	CudaStatus { Default , ForceDisable }
	Shapefinder2 CUDA status enum. More...

Enumeration Type Documentation

ContrastMode

enum ContrastMode

Contrast mode for feature extraction.

Enumerator
Normal	Normal contrast features.
Inverse	Inverted contrast features.

CudaStatus

enum CudaStatus

Shapefinder2 CUDA status enum.

Enumerator
Default	Use CUDA if available
ForceDisable	User disabled CUDA

PrecisionMode

enum PrecisionMode

Search mode to be used.

Enumerator
NoCorrelation	In the NoCorrelation mode, only the ShapeFinder edge model will be searched. Not subsequent additional correlation step will be carried out to improve result accuracy. Results in this mode are pixel- accurate only. This search mode is the fastest available.
CorrelationCoarse	In the CorrelationCoarse mode, after the initial edge model search a correlation and hill climbing will be performed on the preliminary result using the coarse layer of the model. This usually improves the overall result quality and allows for sub-pixel accurate results. As the correlations will be calculate on the coarse layer this mode is a trade-off between speed and accuracy.
CorrelationFine	In the CorrelationFine mode, after the initial edge model search a correlation and hill climbing will be performed on the preliminary result using the fine layer of the model. This usually improves the overall result quality and allows for sub-pixel accurate results.