The MyCaffeControl is the main object used to manage all training, testing and running of the MyCaffe system. More...

Inheritance diagram for MyCaffe.MyCaffeControl< T >:

Public Member Functions
	MyCaffeControl (SettingsCaffe settings, Log log, CancelEvent evtCancel, AutoResetEvent evtSnapshot=null, AutoResetEvent evtForceTest=null, ManualResetEvent evtPause=null, List< int > rgGpuId=null, string strCudaPath="", bool bCreateCudaDnn=false, ConnectInfo ci=null)
	The MyCaffeControl constructor. More...

void	dispose ()
	Releases all GPU and Host resources used by the CaffeControl. More...

MyCaffeControl< T >	Clone (int nGpuID)
	Clone the current instance of the MyCaffeControl creating a second instance. More...

void	CopyGradientsFrom (MyCaffeControl< T > src)
	Copy the learnable parameter diffs from the source MyCaffeControl into this one. More...

void	CopyWeightsFrom (MyCaffeControl< T > src)
	Copy the learnable parameter data from the source MyCaffeControl into this one. More...

double	ApplyUpdate (int nIteration)
	Directs the solver to apply the leanred blob diffs to the weights using the solver's learning rate and update algorithm. More...

void	Unload (bool bUnloadImageDb=true, bool bIgnoreExceptions=false)
	Unload the currently loaded project, if any. More...

bool	ReInitializeParameters (WEIGHT_TARGET target, params string[] rgstrLayers)
	Re-initializes each of the specified layers by re-running the filler (if any) specified by the layer. When the 'rgstr' parameter is null or otherwise empty, the blobs of all layers are re-initialized. More...

void	SetOnTestOverride (EventHandler< TestArgs > onTest)
	Sets the root solver's onTest event function. More...

void	SetOnTrainingStartOverride (EventHandler onTrainingStart)
	Sets the root solver's onStart event function triggered on the start of each training pass. More...

void	SetOnTestingStartOverride (EventHandler onTestingStart)
	Sets the root solver's onTestingStart event function triggered on the start of each testing pass. More...

void	AddCancelOverrideByName (string strEvtCancel)
	Adds a cancel override. More...

void	AddCancelOverride (CancelEvent evtCancel)
	Adds a cancel override. More...

void	RemoveCancelOverrideByName (string strEvtCancel)
	Remove a cancel override. More...

void	RemoveCancelOverride (CancelEvent evtCancel)
	Remove a cancel override. More...

int	GetDeviceCount ()
	Returns the total number of devices installed on this computer. More...

string	GetDeviceName (int nDeviceID)
	Returns the device name of a given device ID. More...

void	PrepareImageMeans (ProjectEx prj)
	Prepare the testing image mean by copying the training image mean if the testing image mean is missing. More...

bool	Load (Phase phase, ProjectEx p, DB_LABEL_SELECTION_METHOD? labelSelectionOverride=null, DB_ITEM_SELECTION_METHOD? itemSelectionOverride=null, bool bResetFirst=false, IXDatabaseBase db=null, bool bUseDb=true, bool bCreateRunNet=true, string strStage=null, bool bEnableMemTrace=false)
	Load a project and optionally the MyCaffeImageDatabase. More...

bool	Load (Phase phase, string strSolver, string strModel, byte[] rgWeights, DB_LABEL_SELECTION_METHOD? labelSelectionOverride=null, DB_ITEM_SELECTION_METHOD? itemSelectionOverride=null, bool bResetFirst=false, IXDatabaseBase db=null, bool bUseDb=true, bool bCreateRunNet=true, string strStage=null, bool bEnableMemTrace=false)
	Load a project and optionally the MyCaffeImageDatabase. More...

bool	LoadLite (Phase phase, string strSolver, string strModel, byte[] rgWeights=null, bool bResetFirst=false, bool bCreateRunNet=true, SimpleDatum sdMean=null, string strStage=null, bool bEnableMemTrace=false)
	Load a solver and model without using the MyCaffeImageDatabase. More...

void	LoadToRun (string strModel, byte[] rgWeights, BlobShape shape, SimpleDatum sdMean=null, TransformationParameter transParam=null, bool bForceBackward=false, bool bConvertToRunNet=true)
	The LoadToRun method loads the MyCaffeControl for running only (e.g. deployment). More...

bool	CompareWeights (Net< T > net1, Net< T > net2)
	The CompareWeights method compares the weights held in two different Net objects. More...

void	Train (int nIterationOverride=-1, int nTrainingTimeLimitInMinutes=0, TRAIN_STEP step=TRAIN_STEP.NONE, double dfLearningRateOverride=0, bool bReset=false)
	Train the network a set number of iterations. More...

double	Test (int nIterationOverride=-1)
	Test the network a given number of iterations. More...

PropertySet	TestMany (PropertySet customInput)
	Test on custom input data. More...

BlobCollection< T >	TestManyEx (PropertySet customInput)
	Test on custom input data. More...

List< Tuple< SimpleDatum, ResultCollection > >	TestMany (int nCount, bool bOnTrainingSet, bool bOnTargetSet=false, DB_ITEM_SELECTION_METHOD imgSelMethod=DB_ITEM_SELECTION_METHOD.RANDOM, int nImageStartIdx=0, DateTime? dtImageStartTime=null, double? dfThreshold=null)
	Test on a number of images by selecting random images from the database, running them through the Run network, and then comparing the results with the expected results. More...

ResultCollection	Run (int nImageIdx, bool bPad=true)
	Run on a given image in the MyCaffeImageDatabase based on its image index. More...

List< ResultCollection >	Run (List< int > rgImageIdx, ref Blob< T > blob)
	Run on a set of images in the MyCaffeImageDatabase based on their image indexes. More...

List< ResultCollection >	Run (List< int > rgImageIdx)
	Run on a set of images in the MyCaffeImageDatabase based on their image indexes. More...

Blob< T >	CreateDataBlob (SimpleDatum d, Blob< T > blob=null, bool bPad=true)
	Create a data blob from a SimpleDatum by transforming the data and placing the results in the blob returned. More...

ResultCollection	Run (SimpleDatum d, bool bSort, bool bUseSolverNet, bool bPad=true)
	Run on a given Datum. More...

List< ResultCollection >	Run (List< SimpleDatum > rgSd, ref Blob< T > blob, bool bUseSolverNet=false, int nMax=int.MaxValue)
	Run on a given list of Datum. More...

List< ResultCollection >	Run (Blob< T > blob, bool bSort=true, bool bUseSolverNet=false, int nMax=int.MaxValue, List< int > rgIgnoreLabels=null)
	Run on a Blob of data. More...

ResultCollection	Run (Bitmap img, bool bSort=true, bool bPad=true)
	Run on a given bitmap image. More...

ResultCollection	Run (SimpleDatum d, bool bSort=true, bool bPad=true)
	Run on a given Datum. More...

PropertySet	RunModel (PropertySet customInput)
	Run the model using data from the model itself - requires a Data layer with the RUN phase. More...

BlobCollection< T >	RunModelEx (PropertySet customInput)
	Run the model using data from the model itself - requires a Data layer with the RUN phase. More...

PropertySet	Run (PropertySet customInput, int nK=1, double dfThreshold=0.01, int nMax=80, bool bBeamSearch=false)
	Run the model on custom input data. More...

BlobCollection< T >	Run (BlobCollection< T > colBottom)
	Run the network forward on the bottom blobs. More...

Bitmap	GetTestImage (Phase phase, out int nLabel, out string strLabel)
	Retrieves a random image from either the training or test set depending on the Phase specified. More...

Bitmap	GetTestImage (Phase phase, int nLabel)
	Retrieves a random image from either the training or test set depending on the Phase specified. More...

Bitmap	GetTargetImage (int nSrcId, int nIdx, out int nLabel, out string strLabel, out byte[] rgCriteria, out SimpleDatum.DATA_FORMAT fmtCriteria)
	Retrives the image at a given index within the Testing data set. More...

Bitmap	GetTargetImage (int nImageID, out int nLabel, out string strLabel, out byte[] rgCriteria, out SimpleDatum.DATA_FORMAT fmtCriteria)
	Retrives the image with a given ID. More...

SimpleDatum	GetItemMean ()
	Returns the item (e.g., image or temporal item) mean used by the solver network used during training. More...

DatasetDescriptor	GetDataset ()
	Returns the current dataset used when training and testing. More...

byte[]	GetWeights ()
	Retrieves the weights of the training network. More...

void	UpdateRunWeights (bool bOutputStatus=false, bool bVerifyWeights=true)
	Loads the weights from the training net into the Net used for running. More...

void	UpdateWeights (byte[] rgWeights)
	Loads the training Net with new weights. More...

Net< T >	CreateNet (byte[] rgWeights, CudaDnn< T > cudaOverride=null)
	Creates a new Net, loads the weights specified into it and returns it. More...

Net< T >	GetInternalNet (Phase phase=Phase.RUN)
	Returns the internal net based on the Phase specified: TRAIN, TEST or RUN. More...

Solver< T >	GetInternalSolver ()
	Get the internal solver. More...

void	Snapshot (bool bUpdateDatabase=true)
	The Snapshot function forces a snapshot to occur. More...

string	GetLicenseText (string strOtherLicenses)
	Returns the license text for MyCaffe. More...

bool	VerifyCompute (string strExtra=null, int nDeviceID=-1, bool bThrowException=true)
	VerifyCompute compares the current compute of the current device (or device specified) against the required compute of the current CudaDnnDLL.dll used. More...

Blob< T >	CreateBlob (string strName)
	Create an unsized blob and set its name. More...

long	CreateExtension (string strExtensionDLLPath)
	Create and load a new extension DLL. More...

void	FreeExtension (long hExtension)
	Free an existing extension and unload it. More...

T[]	RunExtension (long hExtension, long lfnIdx, T[] rgParam)
	Run a function on an existing extension. More...

double[]	RunExtensionD (long hExtension, long lfnIdx, double[] rgParam)
	Run a function on an existing extension using the double base type. More...

float[]	RunExtensionF (long hExtension, long lfnIdx, float[] rgParam)
	Run a function on an existing extension using the float base type. More...

Public Member Functions inherited from System.ComponentModel.Component
	Component ()

Static Public Member Functions
static NetParameter	CreateNetParameterForRunning (BlobShape shape, string strModel, out TransformationParameter transform_param, Stage stage=Stage.NONE, bool bSkipLossLayer=false, bool bMaintainBatchSize=false)
	Creates a net parameter for the RUN phase. More...

static void	ResetDevice (int nDeviceID)
	Reset the device at the given device ID. More...

static string	GetLicenseTextEx (string strOtherLicenses)
	Returns the license text for MyCaffe. More...

Protected Member Functions
NetParameter	createNetParameterForRunning (ProjectEx p, out TransformationParameter transform_param)
	Creates a net parameter for the RUN phase. More...

NetParameter	createNetParameterForRunning (DatasetDescriptor ds, string strModel, out TransformationParameter transform_param, Stage stage=Stage.NONE)
	Creates a net parameter for the RUN phase. More...

NetParameter	createNetParameterForRunning (BlobShape shape, string strModel, out TransformationParameter transform_param, Stage stage=Stage.NONE)
	Creates a net parameter for the RUN phase. More...

NetParameter	createNetParameterForRunning (SimpleDatum sdMean, string strModel, out TransformationParameter transform_param, out int nC, out int nH, out int nW, Stage stage=Stage.NONE)
	Creates a net parameter for the RUN phase. More...

Protected Attributes
SettingsCaffe	m_settings
	The settings used to configure the control. More...

Log	m_log
	The log used for output. More...

IXDatabaseBase	m_db = null
	The image database. More...

bool	m_bDbOwner = true
	Whether or not the control owns the image database. More...

CancelEvent	m_evtCancel
	The CancelEvent used to cancel training and testing operations. More...

AutoResetEvent	m_evtForceSnapshot
	An auto-reset event used to force a snapshot. More...

AutoResetEvent	m_evtForceTest
	An auto-reset event used to force a test cycle. More...

ManualResetEvent	m_evtPause
	An auto-reset event used to pause training. More...

DataTransformer< T >	m_dataTransformer = null
	The data transformer used to transform data. More...

ProjectEx	m_project = null
	The active project (if any). More...

DatasetDescriptor	m_dataSet = null
	The dataset descriptor of the dataset used in the image database. More...

string	m_strCudaPath = null
	The low-level path of the underlying CudaDnn DLL. More...

List< int >	m_rgGpu
	A list of the Device ID's used for training. More...

Properties
static FileVersionInfo	Version `[get]`
	Get the file version of the MyCaffe assembly running. More...

ConnectInfo	DatasetConnectInfo `[get]`
	Returns the dataset connection information, if used (default = null). More...

string	CurrentStage `[get]`
	Returns the stage under which the project was loaded, if any. More...

bool	EnableTesting `[getset]`
	Enable/disable testing. For example reinforcement learning does not use testing. More...

bool	EnableVerboseStatus `[getset]`
	Get/set whether or not to use verbose status. When enabled, the full status is output when loading a project, otherwise a more minimum (faster) set is output (default = false for disabled). More...

bool?	EnableBlobDebugging `[getset]`
	Enable/disable blob debugging. More...

bool?	EnableBreakOnFirstNaN `[getset]`
	Enable/disable break training after first detecting a NaN. More...

bool?	EnableDetailedNanDetection `[getset]`
	When enabled (requires EnableBlobDebugging = true), the detailed Nan (and Infinity) detection is perofmed on each blob when training Net. More...

bool?	EnableLayerDebugging `[getset]`
	Enable/disable layer debugging which causes each layer to check for NAN/INF on each forward/backward pass and throw an exception when found. More...

bool?	EnableSingleStep `[getset]`
	Enable/disable single step training. More...

DataTransformer< T >	DataTransformer `[get]`
	Returns the DataTransormer used. More...

SettingsCaffe	Settings `[get]`
	Returns the settings used to create the control. More...

CudaDnn< T >	Cuda `[get]`
	Returns the CudaDnn connection used. More...

Log	Log `[get]`
	Returns the Log (for output) used. More...

IXPersist< T >	Persist `[get]`
	Returns the persist used to load and save weights. More...

IXDatabaseBase	Database `[get]`
	Returns the MyCaffeImageDatabase or MyCaffeTemporalDatabase used. More...

CancelEvent	CancelEvent `[get]`
	Returns the CancelEvent used. More...

List< int >	ActiveGpus `[get]`
	Returns a list of Active GPU's used by the control. More...

string	ActiveLabelCounts `[get]`
	Returns a string describing the active label counts observed during training. More...

string	LabelQueryHitPercents `[get]`
	Returns a string describing the label query hit percentages observed during training. More...

string	LabelQueryEpochs `[get]`
	Returns a string describing the label query epochs observed during training. More...

string	CurrentDevice `[get]`
	Returns the name of the current device used. More...

ProjectEx	CurrentProject `[get]`
	Returns the name of the currently loaded project. More...

int	CurrentIteration `[get]`
	Returns the current iteration. More...

int	MaximumIteration `[get]`
	Returns the maximum iteration. More...

Phase	LastPhase `[get]`
	Returns the last phase run (TRAIN, TEST or RUN). More...

Properties inherited from MyCaffe.common.IXMyCaffeState< T >
bool	EnableBlobDebugging `[getset]`
	Enable/disable blob debugging. More...

bool	EnableBreakOnFirstNaN `[getset]`
	Enable/disable break training after first detecting a NaN. More...

bool	EnableDetailedNanDetection `[getset]`
	When enabled (requires EnableBlobDebugging = true), the detailed Nan (and Infinity) detection is perofmed on each blob when training Net. More...

bool	EnableSingleStep `[getset]`
	Enable/disable single step training. More...

bool	EnableLayerDebugging `[getset]`
	Enable/disable layer debugging which causes each layer to check for NAN/INF on each forward/backward pass and throw an exception when found. More...

IXPersist< T >	Persist `[get]`
	Returns the persist used to load and save weights. More...

IXDatabaseBase	Database `[get]`
	Returns the in-memory MyCaffeDatabase used. More...

CancelEvent	CancelEvent `[get]`
	Returns the CancelEvent used. More...

List< int >	ActiveGpus `[get]`
	Returns a list of Active GPU's used by the control. More...

string	ActiveLabelCounts `[get]`
	Returns a string describing the active label counts observed during training. More...

string	LabelQueryHitPercents `[get]`
	Returns a string describing the label query hit percentages observed during training. More...

string	LabelQueryEpochs `[get]`
	Returns a string describing the label query epochs observed during training. More...

string	CurrentDevice `[get]`
	Returns the name of the current device used. More...

ProjectEx	CurrentProject `[get]`
	Returns the name of the currently loaded project. More...

int	CurrentIteration `[get]`
	Returns the current iteration. More...

int	MaximumIteration `[get]`
	Returns the maximum iteration. More...

Events
EventHandler< SnapshotArgs >	OnSnapshot
	The OnSnapshot event fires each time a snap-shot is taken. More...

EventHandler< TrainingIterationArgs< T > >	OnTrainingIteration
	The OnTrainingIteration event fires at the end of each training iteration. More...

EventHandler< TestingIterationArgs< T > >	OnTestingIteration
	The OnTestingIteration event fires at the end of each testing iteration. More...

Detailed Description

The MyCaffeControl is the main object used to manage all training, testing and running of the MyCaffe system.

Template Parameters

T	Specifies the base type float or double. Using float is recommended to conserve GPU memory.

Definition at line 34 of file MyCaffeControl.cs.

Constructor & Destructor Documentation

◆ MyCaffeControl()

MyCaffe.MyCaffeControl< T >.MyCaffeControl	(	SettingsCaffe	settings,
		Log	log,
		CancelEvent	evtCancel,
		AutoResetEvent	evtSnapshot = `null`,
		AutoResetEvent	evtForceTest = `null`,
		ManualResetEvent	evtPause = `null`,
		List< int >	rgGpuId = `null`,
		string	strCudaPath = `""`,
		bool	bCreateCudaDnn = `false`,
		ConnectInfo	ci = `null`
	)

The MyCaffeControl constructor.

Parameters

settings	Specifies the setting used to configure the MyCaffeControl.
log	Specifies the Log for all output.
evtCancel	Specifies the CancelEvent used to abort training and testing operations.
evtSnapshot	Optionally, specifies an auto reset event used to force a snap-shot.
evtForceTest	Optionally, specifies an auto reset event used to force a test cycle.
evtPause	Optionally, specifies an auto reset event used to pause training.
rgGpuId	Optionally, specfies a set of GPU ID's that override those specified in the SettingsCaffe object.
strCudaPath	Optionally, specifies the path to the low-lever CudaDnnDll.DLL file. Note, when not set, the system looks in the same directory of the executing assembly for the low-level DLL.
bCreateCudaDnn	Optionally, specififies create the connection to CUDA (default = false, causing the creation to occur during Load).
ci	Optionally, specifies the connection information used to connect to the dataset.

Definition at line 136 of file MyCaffeControl.cs.

Member Function Documentation

◆ AddCancelOverride()

void MyCaffe.MyCaffeControl< T >.AddCancelOverride ( CancelEvent evtCancel )

Adds a cancel override.

Parameters

evtCancel Specifies the new name of the cancel event to add.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 496 of file MyCaffeControl.cs.

◆ AddCancelOverrideByName()

void MyCaffe.MyCaffeControl< T >.AddCancelOverrideByName ( string strEvtCancel )

Adds a cancel override.

Parameters

strEvtCancel Specifies the new name of the cancel event to add.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 487 of file MyCaffeControl.cs.

◆ ApplyUpdate()

double MyCaffe.MyCaffeControl< T >.ApplyUpdate ( int nIteration )

Directs the solver to apply the leanred blob diffs to the weights using the solver's learning rate and update algorithm.

Parameters

nIteration Specifies the current iteration.

Returns: The learning rate used is returned.

Definition at line 359 of file MyCaffeControl.cs.

◆ Clone()

MyCaffeControl< T > MyCaffe.MyCaffeControl< T >.Clone ( int nGpuID )

Clone the current instance of the MyCaffeControl creating a second instance.

The second instance has the same project loaded and a copy of the first instance's weights.

Parameters

nGpuID Specifies the GPUID on which to load the second instance.

Returns: The new MyCaffeControl instance is returned.

Definition at line 285 of file MyCaffeControl.cs.

◆ CompareWeights()

bool MyCaffe.MyCaffeControl< T >.CompareWeights	(	Net< T >	net1,
		Net< T >	net2
	)

The CompareWeights method compares the weights held in two different Net objects.

Parameters

net1	Specifies the first Net to compare.
net2	Specifies the second Net to compare.

Returns: If the weights are identical, true is returned, otherwise false.

Definition at line 1800 of file MyCaffeControl.cs.

◆ CopyGradientsFrom()

void MyCaffe.MyCaffeControl< T >.CopyGradientsFrom ( MyCaffeControl< T > src )

Copy the learnable parameter diffs from the source MyCaffeControl into this one.

Parameters

src	Specifies the source MyCaffeControl whos gradients (blob diffs) are to be copied.

Definition at line 317 of file MyCaffeControl.cs.

◆ CopyWeightsFrom()

void MyCaffe.MyCaffeControl< T >.CopyWeightsFrom ( MyCaffeControl< T > src )

Copy the learnable parameter data from the source MyCaffeControl into this one.

Parameters

src	Specifies the source MyCaffeControl whos gradients (blob data) are to be copied.

Definition at line 337 of file MyCaffeControl.cs.

◆ CreateBlob()

Blob< T > MyCaffe.MyCaffeControl< T >.CreateBlob ( string strName )

Create an unsized blob and set its name.

Parameters

strName Specifies the Blob name.

Returns: The Blob is returned.

Definition at line 3496 of file MyCaffeControl.cs.

◆ CreateDataBlob()

Blob< T > MyCaffe.MyCaffeControl< T >.CreateDataBlob	(	SimpleDatum	d,
		Blob< T >	blob = `null`,
		bool	bPad = `true`
	)

Create a data blob from a SimpleDatum by transforming the data and placing the results in the blob returned.

Parameters

d	Specifies the datum to load into the blob.
blob	Optionally, specifies a blob to use instead of creating a new one.
bPad	Optionally, pad the blob with an extra item (default = false).

Returns: The data blob containing the transformed data is returned.

Implements MyCaffe.common.IXMyCaffeNoDb< T >.

Definition at line 2460 of file MyCaffeControl.cs.

◆ CreateExtension()

long MyCaffe.MyCaffeControl< T >.CreateExtension ( string strExtensionDLLPath )

Create and load a new extension DLL.

Parameters

strExtensionDLLPath Specifies the path to the extension DLL.

Returns: The handle to the extension is returned.

Implements MyCaffe.common.IXMyCaffeExtension< T >.

Definition at line 3508 of file MyCaffeControl.cs.

◆ CreateNet()

Net< T > MyCaffe.MyCaffeControl< T >.CreateNet	(	byte[]	rgWeights,
		CudaDnn< T >	cudaOverride = `null`
	)

Creates a new Net, loads the weights specified into it and returns it.

Parameters

rgWeights	Specifies the weights to load.
cudaOverride	Optionally, specifies a different cuda instance for the Net to use.

Returns: The new Net is returned.

Definition at line 3304 of file MyCaffeControl.cs.

◆ createNetParameterForRunning() [1/4]

NetParameter MyCaffe.MyCaffeControl< T >.createNetParameterForRunning	(	BlobShape	shape,
		string	strModel,
		out TransformationParameter	transform_param,
		Stage	stage = `Stage.NONE`
	)

protected

Creates a net parameter for the RUN phase.

This function transforms the training net parameter into a new net parameter suitable to run in the RUN phase.

Parameters

shape	Specifies the shape of the images that will be used.
strModel	Specifies the model descriptor.
transform_param	Specifies the TransformationParameter to use.
stage	Optionally, specifies the stage to create the run network on.

Returns: The new NetParameter suitable for the RUN phase is returned.

Definition at line 798 of file MyCaffeControl.cs.

◆ CreateNetParameterForRunning()

static NetParameter MyCaffe.MyCaffeControl< T >.CreateNetParameterForRunning	(	BlobShape	shape,
		string	strModel,
		out TransformationParameter	transform_param,
		Stage	stage = `Stage.NONE`,
		bool	bSkipLossLayer = `false`,
		bool	bMaintainBatchSize = `false`
	)

static

Creates a net parameter for the RUN phase.

This function transforms the training net parameter into a new net parameter suitable to run in the RUN phase.

Parameters

shape	Specifies the shape of the images that will be used.
strModel	Specifies the model descriptor.
transform_param	Specifies the TransformationParameter to use.
stage	Optionally, specifies the stage to create the run network on.
bSkipLossLayer	Optionally, specifies to skip the loss layer and not output a converted layer to replace it (default = false).
bMaintainBatchSize	Optionally, specifies to keep the batch size, otherwise batch size is set to 1 (default = false).

Returns: The new NetParameter suitable for the RUN phase is returned.

Definition at line 882 of file MyCaffeControl.cs.

◆ createNetParameterForRunning() [2/4]

NetParameter MyCaffe.MyCaffeControl< T >.createNetParameterForRunning	(	DatasetDescriptor	ds,
		string	strModel,
		out TransformationParameter	transform_param,
		Stage	stage = `Stage.NONE`
	)

protected

Creates a net parameter for the RUN phase.

This function transforms the training net parameter into a new net parameter suitable to run in the RUN phase.

Parameters

ds	Specifies a DatasetDescriptor for the dataset used.
strModel	Specifies the model descriptor.
transform_param	Specifies the TransformationParameter to use.
stage	Optionally, specifies the stage to create the run network on.

Returns: The new NetParameter suitable for the RUN phase is returned.

Definition at line 781 of file MyCaffeControl.cs.

◆ createNetParameterForRunning() [3/4]

NetParameter MyCaffe.MyCaffeControl< T >.createNetParameterForRunning	(	ProjectEx	p,
		out TransformationParameter	transform_param
	)

protected

Creates a net parameter for the RUN phase.

This function transforms the training net parameter into a new net parameter suitable to run in the RUN phase.

Parameters

p	Specifies a project.
transform_param	Specifies the TransformationParameter to use.

Returns: The new NetParameter suitable for the RUN phase is returned.

Definition at line 765 of file MyCaffeControl.cs.

◆ createNetParameterForRunning() [4/4]

NetParameter MyCaffe.MyCaffeControl< T >.createNetParameterForRunning	(	SimpleDatum	sdMean,
		string	strModel,
		out TransformationParameter	transform_param,
		out int	nC,
		out int	nH,
		out int	nW,
		Stage	stage = `Stage.NONE`
	)

protected

Creates a net parameter for the RUN phase.

This function transforms the training net parameter into a new net parameter suitable to run in the RUN phase.

Parameters

sdMean	Specifies the mean image data used to size the network and as the mean image when used in the transformation parameter.
strModel	Specifies the model descriptor.
transform_param	Returns the TransformationParameter to use.
nC	Returns the discovered channel sizing to use.
nH	Returns the discovered height sizing to use.
nW	Returns the discovered width sizing to use.
stage	Optionally, specifies the stage to create the run network on.

Returns: The new NetParameter suitable for the RUN phase is returned.

Definition at line 819 of file MyCaffeControl.cs.

◆ dispose()

void MyCaffe.MyCaffeControl< T >.dispose ( )

Releases all GPU and Host resources used by the CaffeControl.

Definition at line 191 of file MyCaffeControl.cs.

◆ FreeExtension()

void MyCaffe.MyCaffeControl< T >.FreeExtension ( long hExtension )

Free an existing extension and unload it.

Parameters

hExtension Specifies the handle to the extension to free.

Implements MyCaffe.common.IXMyCaffeExtension< T >.

Definition at line 3517 of file MyCaffeControl.cs.

◆ GetDataset()

DatasetDescriptor MyCaffe.MyCaffeControl< T >.GetDataset ( )

Returns the current dataset used when training and testing.

Returns: The DatasetDescriptor is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3179 of file MyCaffeControl.cs.

◆ GetDeviceCount()

int MyCaffe.MyCaffeControl< T >.GetDeviceCount ( )

Returns the total number of devices installed on this computer.

Returns

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 733 of file MyCaffeControl.cs.

◆ GetDeviceName()

string MyCaffe.MyCaffeControl< T >.GetDeviceName ( int nDeviceID )

Returns the device name of a given device ID.

Parameters

nDeviceID Specifies the device ID.

Returns

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 743 of file MyCaffeControl.cs.

◆ GetInternalNet()

Net< T > MyCaffe.MyCaffeControl< T >.GetInternalNet ( Phase phase = Phase.RUN )

Returns the internal net based on the Phase specified: TRAIN, TEST or RUN.

Parameters

phase Specifies the Phase used to select the Net.

Returns: The internal Net is returned.

The following net is returned under the following conditions: phase = ALL, return the net from the LastPhase run. If the LastPhase run = NONE, return the RUN net. phase = n/a, return the default RUN net. phase = NONE, return the default RUN net. phase = TRAIN, return the training net. phase = TEST, return the testing net.

Definition at line 3328 of file MyCaffeControl.cs.

◆ GetInternalSolver()

Solver< T > MyCaffe.MyCaffeControl< T >.GetInternalSolver ( )

Get the internal solver.

Returns

Definition at line 3349 of file MyCaffeControl.cs.

◆ GetItemMean()

SimpleDatum MyCaffe.MyCaffeControl< T >.GetItemMean ( )

Returns the item (e.g., image or temporal item) mean used by the solver network used during training.

Returns: The image mean is returned as a SimpleDatum.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3158 of file MyCaffeControl.cs.

◆ GetLicenseText()

string MyCaffe.MyCaffeControl< T >.GetLicenseText ( string strOtherLicenses )

Returns the license text for MyCaffe.

Parameters

strOtherLicenses Specifies other licenses to append to the license text.

Returns

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3401 of file MyCaffeControl.cs.

◆ GetLicenseTextEx()

static string MyCaffe.MyCaffeControl< T >.GetLicenseTextEx ( string strOtherLicenses )

static

Returns the license text for MyCaffe.

Parameters

strOtherLicenses Specifies other licenses to append to the license text.

Returns

Definition at line 3380 of file MyCaffeControl.cs.

◆ GetTargetImage() [1/2]

Bitmap MyCaffe.MyCaffeControl< T >.GetTargetImage	(	int	nImageID,
		out int	nLabel,
		out string	strLabel,
		out byte[]	rgCriteria,
		out SimpleDatum.DATA_FORMAT	fmtCriteria
	)

Retrives the image with a given ID.

Parameters

nImageID	Specifies the Raw Image ID.
nLabel	Returns the expected label for the image.
strLabel	Returns the expected label name for the image.
rgCriteria	Returns the data criteria if one exists.
fmtCriteria	Returns the format of the data criteria, if one exists.

Returns: The image queried is returned.

Definition at line 3135 of file MyCaffeControl.cs.

◆ GetTargetImage() [2/2]

Bitmap MyCaffe.MyCaffeControl< T >.GetTargetImage	(	int	nSrcId,
		int	nIdx,
		out int	nLabel,
		out string	strLabel,
		out byte[]	rgCriteria,
		out SimpleDatum.DATA_FORMAT	fmtCriteria
	)

Retrives the image at a given index within the Testing data set.

Parameters

nSrcId	Specifies the Source ID.
nIdx	Specifies the image index.
nLabel	Returns the expected label for the image.
strLabel	Returns the expected label name for the image.
rgCriteria	Returns the data criteria if one exists.
fmtCriteria	Returns the format of the data criteria, if one exists.

Returns: The image queried is returned.

Definition at line 3106 of file MyCaffeControl.cs.

◆ GetTestImage() [1/2]

Bitmap MyCaffe.MyCaffeControl< T >.GetTestImage	(	Phase	phase,
		int	nLabel
	)

Retrieves a random image from either the training or test set depending on the Phase specified.

Parameters

phase	Specifies whether to select images from the training set or testing set.
nLabel	Returns the expected label for the image.

Returns: The image queried is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3084 of file MyCaffeControl.cs.

◆ GetTestImage() [2/2]

Bitmap MyCaffe.MyCaffeControl< T >.GetTestImage	(	Phase	phase,
		out int	nLabel,
		out string	strLabel
	)

Retrieves a random image from either the training or test set depending on the Phase specified.

Parameters

phase	Specifies whether to select images from the training set or testing set.
nLabel	Returns the expected label for the image.
strLabel	Returns the expected label name for the image.

Returns: The image queried is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3060 of file MyCaffeControl.cs.

◆ GetWeights()

byte[] MyCaffe.MyCaffeControl< T >.GetWeights ( )

Retrieves the weights of the training network.

Returns: The weights are returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3188 of file MyCaffeControl.cs.

◆ Load() [1/2]

bool MyCaffe.MyCaffeControl< T >.Load	(	Phase	phase,
		ProjectEx	p,
		DB_LABEL_SELECTION_METHOD?	labelSelectionOverride = `null`,
		DB_ITEM_SELECTION_METHOD?	itemSelectionOverride = `null`,
		bool	bResetFirst = `false`,
		IXDatabaseBase	db = `null`,
		bool	bUseDb = `true`,
		bool	bCreateRunNet = `true`,
		string	strStage = `null`,
		bool	bEnableMemTrace = `false`
	)

Load a project and optionally the MyCaffeImageDatabase.

This load function uses the MyCaffeImageDatabase.

Parameters

phase	Specifies the Phase for which the load should focus.
p	Specifies the Project to load.
labelSelectionOverride	Optionally, specifies the label selection override (overides the label selection in SettingsCaffe). The label selection dictates how the label sets are selected.
itemSelectionOverride	Optionally, specifies the item (e.g., image or temporal item) selection override (overides the item selection in SettingsCaffe). The item selection dictates how the items are selected from each label set.
bResetFirst	Optionally, resets the device before loading. IMPORTANT: this functionality is only recommendned during testing, for resetting the device will throw off all other users of the device.
db	Optionally, specifies the MyCaffeImageDatabase or MyCaffeTemporalDatabase to use. When null, an instance if the in-memory database is created internally.
bUseDb	Optionally, specifies to use the in-memory database based on the DB_VERSION specified in the settings.
bCreateRunNet	Optionally, specifies whether or not to create the Run net.
strStage	Optionally, specifies a stage under which to load the model.
bEnableMemTrace	Optionally, specifies to enable the memory tracing (only available in debug builds).

Returns: If the project is loaded the function returns true, otherwise false is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 1070 of file MyCaffeControl.cs.

◆ Load() [2/2]

bool MyCaffe.MyCaffeControl< T >.Load	(	Phase	phase,
		string	strSolver,
		string	strModel,
		byte[]	rgWeights,
		DB_LABEL_SELECTION_METHOD?	labelSelectionOverride = `null`,
		DB_ITEM_SELECTION_METHOD?	itemSelectionOverride = `null`,
		bool	bResetFirst = `false`,
		IXDatabaseBase	db = `null`,
		bool	bUseDb = `true`,
		bool	bCreateRunNet = `true`,
		string	strStage = `null`,
		bool	bEnableMemTrace = `false`
	)

Load a project and optionally the MyCaffeImageDatabase.

This load function uses the MyCaffeImageDatabase.

Parameters

phase	Specifies the Phase for which the load should focus.
strSolver	Specifies the solver descriptor.
strModel	Specifies the model desciptor.
rgWeights	Optionally, specifies the weights to load, or null to ignore.
labelSelectionOverride	Optionally, specifies the label selection override (overides the label selection in SettingsCaffe). The label selection dictates how the label sets are selected.
itemSelectionOverride	Optionally, specifies the item (e.g., image or temporal item) selection override (overides the item selection in SettingsCaffe). The item selection dictates how the item are selected from each label set.
bResetFirst	Optionally, resets the device before loading. IMPORTANT: this functionality is only recommendned during testing, for resetting the device will throw off all other users of the device.
db	Optionally, specifies the in-memory MyCaffeDatabase to use. When null, an instance if the MyCaffeImageDatabase or MyCaffeTemporalDatabase is created internally depending on the DB_VERSION used.
bUseDb	Optionally, specifies to use the in-memory database based on the DB_VERSION specified in the settings.
bCreateRunNet	Optionally, specifies whether or not to create the Run net (default = true).
strStage	Optionally, specifies a stage under which to load the model.
bEnableMemTrace	Optionally, specifies to enable the memory tracing (only available in debug builds).

Returns: If the project is loaded the function returns true, otherwise false is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 1298 of file MyCaffeControl.cs.

◆ LoadLite()

bool MyCaffe.MyCaffeControl< T >.LoadLite	(	Phase	phase,
		string	strSolver,
		string	strModel,
		byte[]	rgWeights = `null`,
		bool	bResetFirst = `false`,
		bool	bCreateRunNet = `true`,
		SimpleDatum	sdMean = `null`,
		string	strStage = `null`,
		bool	bEnableMemTrace = `false`
	)

Load a solver and model without using the MyCaffeImageDatabase.

This load function is a load lite that does not use the MyCaffeImageDatabase.

Parameters

phase	Specifies the Phase for which the load should focus.
strSolver	Specifies the solver descriptor.
strModel	Specifies the model desciptor.
rgWeights	Optionally, specifies the weights to load, or null to ignore (default = null).
bResetFirst	Optionally, resets the device before loading. IMPORTANT: this functionality is only recommendned during testing, for resetting the device will throw off all other users of the device.
bCreateRunNet	Optionally, specifies whether or not to create the Run net (default = true).
sdMean	Optionally, specifies the image mean to use (default = null).
strStage	Optionally, specifies a stage under which to load the model.
bEnableMemTrace	Optionally, specifies to enable the memory tracing (only available in debug builds).

Returns: If the project is loaded the function returns true, otherwise false is returned.

Definition at line 1512 of file MyCaffeControl.cs.

◆ LoadToRun()

void MyCaffe.MyCaffeControl< T >.LoadToRun	(	string	strModel,
		byte[]	rgWeights,
		BlobShape	shape,
		SimpleDatum	sdMean = `null`,
		TransformationParameter	transParam = `null`,
		bool	bForceBackward = `false`,
		bool	bConvertToRunNet = `true`
	)

The LoadToRun method loads the MyCaffeControl for running only (e.g. deployment).

This method does not use the MyCaffeImageDatabase.

Parameters

strModel	Specifies the model description to load.
rgWeights	Specifies the trained weights to load.
shape	Specifies the expected shape to run on.
sdMean	Optionally, specifies the simple datum mean to subtract from input images that are run.
transParam	Optionally, specifies the TransformationParameter to use. When using a 'deployment' model that has no data layers, you should supply a transformation parameter that matches the transformation used during training.
bForceBackward	Optionally, specifies to force backward propagation in the event that a backward pass is to be run on the Run net - The DeepDraw functionality uses this setting so that it can view what the trained weights actually see.
bConvertToRunNet	When true, the 'strModel' is converted from a training model to a run model, otherwise the model is used unaltered (default = true)

Implements MyCaffe.common.IXMyCaffeNoDb< T >.

Definition at line 1637 of file MyCaffeControl.cs.

◆ PrepareImageMeans()

void MyCaffe.MyCaffeControl< T >.PrepareImageMeans ( ProjectEx prj )

Prepare the testing image mean by copying the training image mean if the testing image mean is missing.

Parameters

prj	Specifies the project whos image mean is to be prepared.

Definition at line 994 of file MyCaffeControl.cs.

◆ ReInitializeParameters()

bool MyCaffe.MyCaffeControl< T >.ReInitializeParameters	(	WEIGHT_TARGET	target,
		params string[]	rgstrLayers
	)

Re-initializes each of the specified layers by re-running the filler (if any) specified by the layer.
When the 'rgstr' parameter is null or otherwise empty, the blobs of all layers are re-initialized.

Parameters

target	Specifies the weights to target (e.g. weights, bias or both).
rgstrLayers	Specifies the layers to reinitialize, when null or empty, all layers are re-initialized

Returns: If a layer is specified and found, true is returned, otherwise false is returned.

This method causes the OnTrainingIteration event to fire with the updated values from the re-init.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 448 of file MyCaffeControl.cs.

◆ RemoveCancelOverride()

void MyCaffe.MyCaffeControl< T >.RemoveCancelOverride ( CancelEvent evtCancel )

Remove a cancel override.

Parameters

evtCancel Specifies the new name of the cancel event to remove.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 514 of file MyCaffeControl.cs.

◆ RemoveCancelOverrideByName()

void MyCaffe.MyCaffeControl< T >.RemoveCancelOverrideByName ( string strEvtCancel )

Remove a cancel override.

Parameters

strEvtCancel Specifies the new name of the cancel event to remove.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 505 of file MyCaffeControl.cs.

◆ ResetDevice()

static void MyCaffe.MyCaffeControl< T >.ResetDevice ( int nDeviceID )

static

Reset the device at the given device ID.

WARNING! It is recommended that this only be used when testing, for calling this will throw off all other users of the device and may cause unpredictable behavior.

Parameters

nDeviceID Specifies the device ID of the device to reset.

Definition at line 3371 of file MyCaffeControl.cs.

◆ Run() [1/10]

ResultCollection MyCaffe.MyCaffeControl< T >.Run	(	Bitmap	img,
		bool	bSort = `true`,
		bool	bPad = `true`
	)

Run on a given bitmap image.

This method does not use the MyCaffeImageDatabase.

Parameters

img	Specifies the input image.
bSort	Specifies whether or not to sort the results.
bPad	Optionally, specifies to pad the input by 1.

Returns: The results of the run are returned.

Implements MyCaffe.common.IXMyCaffeNoDb< T >.

Definition at line 2798 of file MyCaffeControl.cs.

◆ Run() [2/10]

List< ResultCollection > MyCaffe.MyCaffeControl< T >.Run	(	Blob< T >	blob,
		bool	bSort = `true`,
		bool	bUseSolverNet = `false`,
		int	nMax = `int.MaxValue`,
		List< int >	rgIgnoreLabels = `null`
	)

Run on a Blob of data.

Parameters

blob	Specifies the blob of data.
bSort	Optionally, specifies whether or not to sor the results.
bUseSolverNet	Optionally, specifies whether or not to use the training net vs. the run net.
nMax	Optionally, specifies a maximum number of SimpleDatums to process (default = int.MaxValue).
rgIgnoreLabels	Optionally, specifies a set of labels to ignore.

Returns: A list of results of the run are returned.

Definition at line 2680 of file MyCaffeControl.cs.

◆ Run() [3/10]

BlobCollection< T > MyCaffe.MyCaffeControl< T >.Run ( BlobCollection< T > colBottom )

Run the network forward on the bottom blobs.

Parameters

colBottom Specifies the input blobs.

Returns: The top output blobs are returned.

Definition at line 3032 of file MyCaffeControl.cs.

◆ Run() [4/10]

ResultCollection MyCaffe.MyCaffeControl< T >.Run	(	int	nImageIdx,
		bool	bPad = `true`
	)

Run on a given image in the MyCaffeImageDatabase based on its image index.

Parameters

nImageIdx	Specifies the image index.
bPad	PAdd the input with an extra input image.

Returns: The result of the run is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 2386 of file MyCaffeControl.cs.

◆ Run() [5/10]

List< ResultCollection > MyCaffe.MyCaffeControl< T >.Run ( List< int > rgImageIdx )

Run on a set of images in the MyCaffeImageDatabase based on their image indexes.

Parameters

rgImageIdx Specifies a list of image indexes.

Returns: A list of results from the run is returned - one result per image.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 2418 of file MyCaffeControl.cs.

◆ Run() [6/10]

List< ResultCollection > MyCaffe.MyCaffeControl< T >.Run	(	List< int >	rgImageIdx,
		ref Blob< T >	blob
	)

Run on a set of images in the MyCaffeImageDatabase based on their image indexes.

Parameters

rgImageIdx	Specifies a list of image indexes.
blob	Specifies a work blob.

Returns: A list of results from the run is returned - one result per image.

Definition at line 2399 of file MyCaffeControl.cs.

◆ Run() [7/10]

List< ResultCollection > MyCaffe.MyCaffeControl< T >.Run	(	List< SimpleDatum >	rgSd,
		ref Blob< T >	blob,
		bool	bUseSolverNet = `false`,
		int	nMax = `int.MaxValue`
	)

Run on a given list of Datum.

Parameters

rgSd	Specifies the list of Datum to run.
blob	Specifies a work blob.
bUseSolverNet	Optionally, specifies whether or not to use the training net vs. the run net.
nMax	Optionally, specifies a maximum number of SimpleDatums to process (default = int.MaxValue).

Returns: A list of results of the run are returned.

Definition at line 2599 of file MyCaffeControl.cs.

◆ Run() [8/10]

PropertySet MyCaffe.MyCaffeControl< T >.Run	(	PropertySet	customInput,
		int	nK = `1`,
		double	dfThreshold = `0.01`,
		int	nMax = `80`,
		bool	bBeamSearch = `false`
	)

Run the model on custom input data.

Parameters

customInput	Specifies the custom input data.
nK	Optionally, specifies the number of items to use in the search where nK=1 runs a greedy search and any values > 1 run a beam search with that width.
dfThreshold	Specifies the threshold where detected items with probabilities less than the threshold are ignored (default = 0.01).
nMax	Optionally, specifies the maximum number of outputs (default = 80).
bBeamSearch	Optionally, specifies to run a Beam Search.

Returns: The results are returned as in a property set.

Definition at line 2906 of file MyCaffeControl.cs.

◆ Run() [9/10]

ResultCollection MyCaffe.MyCaffeControl< T >.Run	(	SimpleDatum	d,
		bool	bSort,
		bool	bUseSolverNet,
		bool	bPad = `true`
	)

Run on a given Datum.

Parameters

d	Specifies the Datum to run.
bSort	Specifies whether or not to sor the results.
bUseSolverNet	Optionally, specifies whether or not to use the training net vs. the run net.
bPad	Optionally, specifies to pad the data with a dummy trailing item (default = true).

Returns: The results of the run are returned.

Definition at line 2507 of file MyCaffeControl.cs.

◆ Run() [10/10]

ResultCollection MyCaffe.MyCaffeControl< T >.Run	(	SimpleDatum	d,
		bool	bSort = `true`,
		bool	bPad = `true`
	)

Run on a given Datum.

Parameters

d	Specifies the Datum to run.
bSort	Specifies whether or not to sort the results.
bPad	Optionally, specifies to pad the data with a dummy item and reshape the net (default = false).

Returns: The results of the run are returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 2818 of file MyCaffeControl.cs.

◆ RunExtension()

T[] MyCaffe.MyCaffeControl< T >.RunExtension	(	long	hExtension,
		long	lfnIdx,
		T[]	rgParam
	)

Run a function on an existing extension.

Parameters

hExtension	Specifies the extension.
lfnIdx	Specifies the function to run on the extension.
rgParam	Specifies the parameters.

Returns: The return values of the function are returned.

Implements MyCaffe.common.IXMyCaffeExtension< T >.

Definition at line 3528 of file MyCaffeControl.cs.

◆ RunExtensionD()

double[] MyCaffe.MyCaffeControl< T >.RunExtensionD	(	long	hExtension,
		long	lfnIdx,
		double[]	rgParam
	)

Run a function on an existing extension using the double base type.

Parameters

hExtension	Specifies the extension.
lfnIdx	Specifies the function to run on the extension.
rgParam	Specifies the parameters.

Returns: The return values of the function are returned.

Implements MyCaffe.common.IXMyCaffeExtension< T >.

Definition at line 3539 of file MyCaffeControl.cs.

◆ RunExtensionF()

float[] MyCaffe.MyCaffeControl< T >.RunExtensionF	(	long	hExtension,
		long	lfnIdx,
		float[]	rgParam
	)

Run a function on an existing extension using the float base type.

Parameters

hExtension	Specifies the extension.
lfnIdx	Specifies the function to run on the extension.
rgParam	Specifies the parameters.

Returns: The return values of the function are returned.

Implements MyCaffe.common.IXMyCaffeExtension< T >.

Definition at line 3556 of file MyCaffeControl.cs.

◆ RunModel()

PropertySet MyCaffe.MyCaffeControl< T >.RunModel ( PropertySet customInput )

Run the model using data from the model itself - requires a Data layer with the RUN phase.

Parameters

customInput Specifies custom inputs. Properties used: 'Phase' specifies the phase for which to run the model (e.g. which net to use), when missing, the default run net is used.

Returns: The results are returned in a property set, where each blob is stored as a byte array packed with the 'float' values from each blob.

Definition at line 2828 of file MyCaffeControl.cs.

◆ RunModelEx()

BlobCollection< T > MyCaffe.MyCaffeControl< T >.RunModelEx ( PropertySet customInput )

Run the model using data from the model itself - requires a Data layer with the RUN phase.

Parameters

customInput Specifies custom inputs. Properties used: 'Phase' specifies the phase for which to run the model (e.g. which net to use), when missing, the default run net is used.

Returns: The results are returned in a property set, where each blob is stored as a byte array packed with the 'float' values from each blob.

Definition at line 2872 of file MyCaffeControl.cs.

◆ SetOnTestingStartOverride()

void MyCaffe.MyCaffeControl< T >.SetOnTestingStartOverride ( EventHandler onTestingStart )

Sets the root solver's onTestingStart event function triggered on the start of each testing pass.

Parameters

onTestingStart Specifies the event handler called when testing.

Definition at line 477 of file MyCaffeControl.cs.

◆ SetOnTestOverride()

void MyCaffe.MyCaffeControl< T >.SetOnTestOverride ( EventHandler< TestArgs > onTest )

Sets the root solver's onTest event function.

Parameters

onTest Specifies the event handler called when testing.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 459 of file MyCaffeControl.cs.

◆ SetOnTrainingStartOverride()

void MyCaffe.MyCaffeControl< T >.SetOnTrainingStartOverride ( EventHandler onTrainingStart )

Sets the root solver's onStart event function triggered on the start of each training pass.

Parameters

onTrainingStart Specifies the event handler called when testing.

Definition at line 468 of file MyCaffeControl.cs.

◆ Snapshot()

void MyCaffe.MyCaffeControl< T >.Snapshot ( bool bUpdateDatabase = true )

The Snapshot function forces a snapshot to occur.

Parameters

bUpdateDatabase Optionally, specifies to update the database (default = true).

Definition at line 3358 of file MyCaffeControl.cs.

◆ Test()

double MyCaffe.MyCaffeControl< T >.Test ( int nIterationOverride = -1 )

Test the network a given number of iterations.

Parameters

nIterationOverride Optionally, specifies number of iterations to run that override the iterations specified in the solver desctiptor.

Returns: The accuracy value from the test is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 1942 of file MyCaffeControl.cs.

◆ TestMany() [1/2]

List< Tuple< SimpleDatum, ResultCollection > > MyCaffe.MyCaffeControl< T >.TestMany	(	int	nCount,
		bool	bOnTrainingSet,
		bool	bOnTargetSet = `false`,
		DB_ITEM_SELECTION_METHOD	imgSelMethod = `DB_ITEM_SELECTION_METHOD.RANDOM`,
		int	nImageStartIdx = `0`,
		DateTime?	dtImageStartTime = `null`,
		double?	dfThreshold = `null`
	)

Test on a number of images by selecting random images from the database, running them through the Run network, and then comparing the results with the expected results.

Parameters

nCount	Specifies the number of cycles to run.
bOnTrainingSet	Specifies on whether to select images from the training set, or when false the testing set of data.
bOnTargetSet	Optionally, specifies to test on the target dataset (if exists) as opposed to the source dataset. The default is false, which tests on the default (source) dataset.
imgSelMethod	Optionally, specifies the image selection method (default = RANDOM).
nImageStartIdx	Optionally, specifies the image start index (default = 0).
dtImageStartTime	Optionally, specifies the image start time (default = null). Note either the 'nImageStartIdx' or 'dtImageStartTime' may be used, but not both.
dfThreshold	Optionally, specifies a threshold where the winning selection must also be above the threshold in score.

Returns: The list of SimpleDatum and their ResultCollections (after running the model on each) is returned.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 2023 of file MyCaffeControl.cs.

◆ TestMany() [2/2]

PropertySet MyCaffe.MyCaffeControl< T >.TestMany ( PropertySet customInput )

Test on custom input data.

Parameters

customInput Specifies the custom input data separated by ';' characters.

Returns: A property set containing the results is returned.

Running test many on custom data requires a MODEL dataset, where the data is queried from the model itself.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 1961 of file MyCaffeControl.cs.

◆ TestManyEx()

BlobCollection< T > MyCaffe.MyCaffeControl< T >.TestManyEx ( PropertySet customInput )

Test on custom input data.

Parameters

customInput Specifies the custom input data separated by ';' characters.

Returns: A property set containing the results is returned.

Running test many on custom data requires a MODEL dataset, where the data is queried from the model itself.

Definition at line 1995 of file MyCaffeControl.cs.

◆ Train()

void MyCaffe.MyCaffeControl< T >.Train	(	int	nIterationOverride = `-1`,
		int	nTrainingTimeLimitInMinutes = `0`,
		TRAIN_STEP	step = `TRAIN_STEP.NONE`,
		double	dfLearningRateOverride = `0`,
		bool	bReset = `false`
	)

Train the network a set number of iterations.

Parameters

nIterationOverride	Optionally, specifies number of iterations to run that override the iterations specified in the solver desctiptor.
nTrainingTimeLimitInMinutes	Optionally, specifies a maximum number of minutes to train. When set to 0, this parameter is ignored and no time limit is imposed.
step	Optionally, specifies whether or not to single step the training on the forward pass, backward pass or both. The default is TRAIN_STEP.NONE which runs the training to the maximum number of iterations specified.
dfLearningRateOverride	Optionally, specifies a learning rate override (default = 0 which ignores this parameter)
bReset	Optionally, reset the iterations to zero.

Note when single stepping, no testing cycles are performed. Currently, the single-step parameter is only suppored when running in single GPU mode.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 1876 of file MyCaffeControl.cs.

◆ Unload()

void MyCaffe.MyCaffeControl< T >.Unload	(	bool	bUnloadImageDb = `true`,
		bool	bIgnoreExceptions = `false`
	)

Unload the currently loaded project, if any.

Parameters

bUnloadImageDb	Optionally, specifies to unload the image database (default = true).
bIgnoreExceptions	Optionally, specifies to ignore exceptions on error (default = false).

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 387 of file MyCaffeControl.cs.

◆ UpdateRunWeights()

void MyCaffe.MyCaffeControl< T >.UpdateRunWeights	(	bool	bOutputStatus = `false`,
		bool	bVerifyWeights = `true`
	)

Loads the weights from the training net into the Net used for running.

Parameters

bOutputStatus	Optionally, specifies to output status as the weights are updated (default = false).
bVerifyWeights	Optionally, specifies to verify the run weights copied (default = true).

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3206 of file MyCaffeControl.cs.

◆ UpdateWeights()

void MyCaffe.MyCaffeControl< T >.UpdateWeights ( byte[] rgWeights )

Loads the training Net with new weights.

Parameters

rgWeights Specifies the weights to load.

Implements MyCaffe.common.IXMyCaffe< T >.

Definition at line 3277 of file MyCaffeControl.cs.

◆ VerifyCompute()

bool MyCaffe.MyCaffeControl< T >.VerifyCompute	(	string	strExtra = `null`,
		int	nDeviceID = `-1`,
		bool	bThrowException = `true`
	)

VerifyCompute compares the current compute of the current device (or device specified) against the required compute of the current CudaDnnDLL.dll used.

Parameters

strExtra	Optionally, specifies extra information for the exception if one is thrown.
nDeviceID	Optionally, specifies a specific device ID to check, otherwise uses the current device used (default = -1, which uses the current device).
bThrowException	Optionally, specifies whether or not to throw an exception on a compute mis-match (default = true).

Returns: If the device's compute is >= to the required compute fo the CudaDnnDll.dll used, true is returned, otherwise false is returned.

Implements MyCaffe.common.IXMyCaffeState< T >.

Definition at line 3413 of file MyCaffeControl.cs.

Member Data Documentation

◆ m_bDbOwner

bool MyCaffe.MyCaffeControl< T >.m_bDbOwner = true

protected

Whether or not the control owns the image database.

Definition at line 51 of file MyCaffeControl.cs.

◆ m_dataSet

DatasetDescriptor MyCaffe.MyCaffeControl< T >.m_dataSet = null

protected

The dataset descriptor of the dataset used in the image database.

Definition at line 79 of file MyCaffeControl.cs.

◆ m_dataTransformer

DataTransformer<T> MyCaffe.MyCaffeControl< T >.m_dataTransformer = null

protected

The data transformer used to transform data.

Definition at line 71 of file MyCaffeControl.cs.

◆ m_db

IXDatabaseBase MyCaffe.MyCaffeControl< T >.m_db = null

protected

The image database.

Definition at line 47 of file MyCaffeControl.cs.

◆ m_evtCancel

CancelEvent MyCaffe.MyCaffeControl< T >.m_evtCancel

protected

The CancelEvent used to cancel training and testing operations.

Definition at line 55 of file MyCaffeControl.cs.

◆ m_evtForceSnapshot

AutoResetEvent MyCaffe.MyCaffeControl< T >.m_evtForceSnapshot

protected

An auto-reset event used to force a snapshot.

Definition at line 59 of file MyCaffeControl.cs.

◆ m_evtForceTest

AutoResetEvent MyCaffe.MyCaffeControl< T >.m_evtForceTest

protected

An auto-reset event used to force a test cycle.

Definition at line 63 of file MyCaffeControl.cs.

◆ m_evtPause

ManualResetEvent MyCaffe.MyCaffeControl< T >.m_evtPause

protected

An auto-reset event used to pause training.

Definition at line 67 of file MyCaffeControl.cs.

◆ m_log

Log MyCaffe.MyCaffeControl< T >.m_log

protected

The log used for output.

Definition at line 43 of file MyCaffeControl.cs.

◆ m_project

ProjectEx MyCaffe.MyCaffeControl< T >.m_project = null

protected

The active project (if any).

Definition at line 75 of file MyCaffeControl.cs.

◆ m_rgGpu

List<int> MyCaffe.MyCaffeControl< T >.m_rgGpu

protected

A list of the Device ID's used for training.

Definition at line 87 of file MyCaffeControl.cs.

◆ m_settings

SettingsCaffe MyCaffe.MyCaffeControl< T >.m_settings

protected

The settings used to configure the control.

Definition at line 39 of file MyCaffeControl.cs.

◆ m_strCudaPath

string MyCaffe.MyCaffeControl< T >.m_strCudaPath = null

protected

The low-level path of the underlying CudaDnn DLL.

Definition at line 83 of file MyCaffeControl.cs.

Property Documentation

◆ ActiveGpus

List<int> MyCaffe.MyCaffeControl< T >.ActiveGpus

get

Returns a list of Active GPU's used by the control.

Definition at line 655 of file MyCaffeControl.cs.

◆ ActiveLabelCounts

string MyCaffe.MyCaffeControl< T >.ActiveLabelCounts

get

Returns a string describing the active label counts observed during training.

This string can help diagnose label balancing issue.

Definition at line 666 of file MyCaffeControl.cs.

◆ CancelEvent

CancelEvent MyCaffe.MyCaffeControl< T >.CancelEvent

get

Returns the CancelEvent used.

Definition at line 647 of file MyCaffeControl.cs.

◆ Cuda

CudaDnn<T> MyCaffe.MyCaffeControl< T >.Cuda

get

Returns the CudaDnn connection used.

Definition at line 615 of file MyCaffeControl.cs.

◆ CurrentDevice

string MyCaffe.MyCaffeControl< T >.CurrentDevice

get

Returns the name of the current device used.

Definition at line 696 of file MyCaffeControl.cs.

◆ CurrentIteration

int MyCaffe.MyCaffeControl< T >.CurrentIteration

get

Returns the current iteration.

Definition at line 716 of file MyCaffeControl.cs.

◆ CurrentProject

ProjectEx MyCaffe.MyCaffeControl< T >.CurrentProject

get

Returns the name of the currently loaded project.

Definition at line 708 of file MyCaffeControl.cs.

◆ CurrentStage

string MyCaffe.MyCaffeControl< T >.CurrentStage

get

Returns the stage under which the project was loaded, if any.

Definition at line 272 of file MyCaffeControl.cs.

◆ Database

IXDatabaseBase MyCaffe.MyCaffeControl< T >.Database

get

Returns the MyCaffeImageDatabase or MyCaffeTemporalDatabase used.

Definition at line 639 of file MyCaffeControl.cs.

◆ DatasetConnectInfo

ConnectInfo MyCaffe.MyCaffeControl< T >.DatasetConnectInfo

get

Returns the dataset connection information, if used (default = null).

Definition at line 264 of file MyCaffeControl.cs.

◆ DataTransformer

DataTransformer<T> MyCaffe.MyCaffeControl< T >.DataTransformer

get

Returns the DataTransormer used.

Definition at line 599 of file MyCaffeControl.cs.

◆ EnableBlobDebugging

bool? MyCaffe.MyCaffeControl< T >.EnableBlobDebugging

getset

Enable/disable blob debugging.

Note, when enabled, training will dramatically slow down.

Definition at line 525 of file MyCaffeControl.cs.

◆ EnableBreakOnFirstNaN

bool? MyCaffe.MyCaffeControl< T >.EnableBreakOnFirstNaN

getset

Enable/disable break training after first detecting a NaN.

This option requires that EnableBlobDebugging == true.

Definition at line 541 of file MyCaffeControl.cs.

◆ EnableDetailedNanDetection

bool? MyCaffe.MyCaffeControl< T >.EnableDetailedNanDetection

getset

When enabled (requires EnableBlobDebugging = true), the detailed Nan (and Infinity) detection is perofmed on each blob when training Net.

Definition at line 554 of file MyCaffeControl.cs.

◆ EnableLayerDebugging

bool? MyCaffe.MyCaffeControl< T >.EnableLayerDebugging

getset

Enable/disable layer debugging which causes each layer to check for NAN/INF on each forward/backward pass and throw an exception when found.

This option dramatically slows down training and is only recommended during debugging.

Definition at line 570 of file MyCaffeControl.cs.

◆ EnableSingleStep

bool? MyCaffe.MyCaffeControl< T >.EnableSingleStep

getset

Enable/disable single step training.

This option requires that EnableBlobDebugging == true.

Definition at line 586 of file MyCaffeControl.cs.

◆ EnableTesting

bool MyCaffe.MyCaffeControl< T >.EnableTesting

getset

Enable/disable testing. For example reinforcement learning does not use testing.

Definition at line 367 of file MyCaffeControl.cs.

◆ EnableVerboseStatus

bool MyCaffe.MyCaffeControl< T >.EnableVerboseStatus

getset

Get/set whether or not to use verbose status. When enabled, the full status is output when loading a project, otherwise a more minimum (faster) set is output (default = false for disabled).

Definition at line 376 of file MyCaffeControl.cs.

◆ LabelQueryEpochs

string MyCaffe.MyCaffeControl< T >.LabelQueryEpochs

get

Returns a string describing the label query epochs observed during training.

This string can help diagnose label balancing issue.

Definition at line 688 of file MyCaffeControl.cs.

◆ LabelQueryHitPercents

string MyCaffe.MyCaffeControl< T >.LabelQueryHitPercents

get

Returns a string describing the label query hit percentages observed during training.

This string can help diagnose label balancing issue.

Definition at line 677 of file MyCaffeControl.cs.

◆ LastPhase

Phase MyCaffe.MyCaffeControl< T >.LastPhase

get

Returns the last phase run (TRAIN, TEST or RUN).

Definition at line 751 of file MyCaffeControl.cs.

◆ Log

Log MyCaffe.MyCaffeControl< T >.Log

get

Returns the Log (for output) used.

Definition at line 623 of file MyCaffeControl.cs.

◆ MaximumIteration

int MyCaffe.MyCaffeControl< T >.MaximumIteration

get

Returns the maximum iteration.

Definition at line 724 of file MyCaffeControl.cs.

◆ Persist

IXPersist<T> MyCaffe.MyCaffeControl< T >.Persist

get

Returns the persist used to load and save weights.

Definition at line 631 of file MyCaffeControl.cs.

◆ Settings

SettingsCaffe MyCaffe.MyCaffeControl< T >.Settings

get

Returns the settings used to create the control.

Definition at line 607 of file MyCaffeControl.cs.

◆ Version

FileVersionInfo MyCaffe.MyCaffeControl< T >.Version

staticget

Get the file version of the MyCaffe assembly running.

Definition at line 252 of file MyCaffeControl.cs.

Event Documentation

◆ OnSnapshot

EventHandler<SnapshotArgs> MyCaffe.MyCaffeControl< T >.OnSnapshot

The OnSnapshot event fires each time a snap-shot is taken.

Definition at line 112 of file MyCaffeControl.cs.

◆ OnTestingIteration

EventHandler<TestingIterationArgs<T> > MyCaffe.MyCaffeControl< T >.OnTestingIteration

The OnTestingIteration event fires at the end of each testing iteration.

Definition at line 120 of file MyCaffeControl.cs.

◆ OnTrainingIteration

EventHandler<TrainingIterationArgs<T> > MyCaffe.MyCaffeControl< T >.OnTrainingIteration

The OnTrainingIteration event fires at the end of each training iteration.

Definition at line 116 of file MyCaffeControl.cs.

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

C:/Data/Data/SS_Projects/Intelligence/GitHub/MyCaffe/MyCaffe/MyCaffeControl.cs

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Properties

Events

Detailed Description

Constructor & Destructor Documentation

◆ MyCaffeControl()

Member Function Documentation

◆ AddCancelOverride()

◆ AddCancelOverrideByName()

◆ ApplyUpdate()

◆ Clone()

◆ CompareWeights()

◆ CopyGradientsFrom()

◆ CopyWeightsFrom()

◆ CreateBlob()

◆ CreateDataBlob()

◆ CreateExtension()

◆ CreateNet()

◆ createNetParameterForRunning() [1/4]

◆ CreateNetParameterForRunning()

◆ createNetParameterForRunning() [2/4]

◆ createNetParameterForRunning() [3/4]

◆ createNetParameterForRunning() [4/4]

◆ dispose()

◆ FreeExtension()

◆ GetDataset()

◆ GetDeviceCount()

◆ GetDeviceName()

◆ GetInternalNet()

◆ GetInternalSolver()

◆ GetItemMean()

◆ GetLicenseText()

◆ GetLicenseTextEx()

◆ GetTargetImage() [1/2]

◆ GetTargetImage() [2/2]

◆ GetTestImage() [1/2]

◆ GetTestImage() [2/2]

◆ GetWeights()

◆ Load() [1/2]

◆ Load() [2/2]

◆ LoadLite()

◆ LoadToRun()

◆ PrepareImageMeans()

◆ ReInitializeParameters()

◆ RemoveCancelOverride()

◆ RemoveCancelOverrideByName()

◆ ResetDevice()

◆ Run() [1/10]

◆ Run() [2/10]

◆ Run() [3/10]

◆ Run() [4/10]

◆ Run() [5/10]

◆ Run() [6/10]

◆ Run() [7/10]

◆ Run() [8/10]

◆ Run() [9/10]

◆ Run() [10/10]

◆ RunExtension()

◆ RunExtensionD()

◆ RunExtensionF()

◆ RunModel()

◆ RunModelEx()

◆ SetOnTestingStartOverride()

◆ SetOnTestOverride()

◆ SetOnTrainingStartOverride()

◆ Snapshot()

◆ Test()

◆ TestMany() [1/2]

◆ TestMany() [2/2]

◆ TestManyEx()

◆ Train()

◆ Unload()

◆ UpdateRunWeights()

◆ UpdateWeights()

◆ VerifyCompute()

Member Data Documentation

◆ m_bDbOwner