only include Magick++ if needed More...

#include <ImageDataSet.h>

Inheritance diagram for ImageDataSet:

Public Member Functions
	ImageDataSet (size_t inputCount, size_t outputCount)

virtual void	PrepareDirectory (const char *filePath)

void	Cache ()

Public Member Functions inherited from DataSet
	DataSet (const char *filePath, size_t inputCount, size_t outputCount)

	DataSet (size_t inputCount, size_t outputCount)

virtual void	PrepareDirectory (const char *filePath)

Public Attributes
size_t	imageHeight = 180
	desired image height More...

size_t	imageWidth = 180
	desired image width More...

double	validationShare = 0.2
	percentage of validation data More...

const char *	trainDirectory = "../../resources/image_classification/training/"
	target directory for training More...

std::vector< MatrixDS< bool > >	classes
	representation of all classes More...

std::vector< std::string >	classNames
	representation of all class names More...

size_t	totalCount = 0
	total number of data records More...

size_t	trainingCount = 0
	number of training data records More...

size_t	validationCount = 0
	number of validation data records More...

Public Attributes inherited from DataSet
size_t	InputCount
	number input elements More...

size_t	OutputCount
	number output elements More...

Set	Training
	set for training More...

Set	Validation
	set for validation of training More...

Set	Test
	set to test after training More...

int	maxEpoch = 1000
	number of epochs while training More...

double	stopThreshold = 0.001
	threshold for loss to prevent over-fitting More...

double	eta = 0.0051
	learning rate More...

int	batchSize = 5
	number of elements per batch More...

bool	verbose = false
	use verbose output during fitting More...

Detailed Description

only include Magick++ if needed

Image data set representation

Constructor & Destructor Documentation

◆ ImageDataSet()

ImageDataSet::ImageDataSet	(	size_t	inputCount,
		size_t	outputCount
	)

inline

default constructor

Parameters

inputCount
outputCount

Member Function Documentation

◆ Cache()

void ImageDataSet::Cache ( )

inline

loads data set into memory

we split the dataset into

80% training
20% validation
0% test

Example: 100 images => 80 training, 20 validation, 0 test

◆ PrepareDirectory()

virtual void ImageDataSet::PrepareDirectory ( const char * filePath )

inlinevirtual

prepares a new set of data based on passed filePath

Parameters

filePath path name of directory to prepare

identifies "classes" as files in nested directories

Reimplemented from DataSet.

Member Data Documentation

◆ classes

std::vector<MatrixDS<bool> > ImageDataSet::classes

representation of all classes

◆ classNames

std::vector<std::string> ImageDataSet::classNames

representation of all class names

◆ imageHeight

size_t ImageDataSet::imageHeight = 180

desired image height

◆ imageWidth

size_t ImageDataSet::imageWidth = 180

desired image width

◆ totalCount

size_t ImageDataSet::totalCount = 0

total number of data records

◆ trainDirectory

const char* ImageDataSet::trainDirectory = "../../resources/image_classification/training/"

target directory for training

◆ trainingCount

size_t ImageDataSet::trainingCount = 0

number of training data records

◆ validationCount

size_t ImageDataSet::validationCount = 0

number of validation data records

◆ validationShare

double ImageDataSet::validationShare = 0.2

percentage of validation data

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

include/math/ds/ImageDataSet.h

Public Member Functions

Public Attributes