EDeepLearningTool represents the common operations of deep learning tools. More...

Inheritance diagram for eVision.EasyDeepLearning.EDeepLearningTool:

[legend]

Collaboration diagram for eVision.EasyDeepLearning.EDeepLearningTool:

Public Member Functions
virtual void	Load (string filePath)
	Loads a deep learning tool. The given ESerializer must have been created for reading.
virtual void	Load (eVision.ESerializer serializer)
	Loads a deep learning tool. The given ESerializer must have been created for reading.
eVision.EasyDeepLearning.EDeepLearningTool	Clone ()
	Clones this tool.
virtual void	Save (string filePath, bool includeTrainingModel)
	Saves the settings, inference and training model of the deep learning tool.
virtual void	Save (string filePath)
	Saves the settings, inference and training model of the deep learning tool.
virtual void	Save (eVision.ESerializer serializer, bool includeTrainingModel)
	Saves the settings, inference and training model of the deep learning tool.
virtual void	Save (eVision.ESerializer serializer)
	Saves the settings, inference and training model of the deep learning tool.
void	SerializeSettings (eVision.ESerializer serializer)
	Serializes the tool settings.
void	SerializeInferenceModel (eVision.ESerializer serializer)
	Serializes the inference model of the tool.
void	SerializeTrainingModel (eVision.ESerializer serializer)
	Serializes the training model of the tool.
void	SaveSettings (string filePath)
	Saves the settings of the tool to a file.
void	SaveInferenceModel (string filePath)
	Saves the tool's inference model to a file.
void	SaveTrainingModel (string filePath)
	Saves the tool's training model to a file.
void	LoadSettings (string filePath)
	Loads the settings of a tool.
void	LoadInferenceModel (string filePath)
	Loads the tool's inference model. The inference model is required to apply the tool on new images.
void	LoadTrainingModel (string filePath)
	Loads the tool's training model. The training model is required to continue a training.
void	UnloadInferenceModel ()
	Unloads the inference model.
void	UnloadTrainingModel ()
	Unloads the training model.
void	UnloadModels ()
	Unloads the inference and training model.
bool	HasInferenceModel ()
	Whether the inference model is loaded or created when the tool is not yet trained.
bool	HasTrainingModel ()
	Whether the training model is loaded or created when the tool is not yet trained.
bool	IsTrained ()
	Tells whether the deep learning tool has been trained.
string	GetAvailableDeviceName (int i)
	Name of the i-th device of the current engine.
eVision.EasyDeepLearning.EDeepLearningDevice	GetAvailableDevice (int i)
	i-th EDeepLearningDevice of the current engine.
void	SetActiveDevice (eVision.EasyDeepLearning.EDeepLearningDevice device)
	Sets the device with which to run or train the model.
eVision.EasyDeepLearning.EDeepLearningDevice	GetActiveDevice (int i)
	Gets the active device. By default, it returns the first active device.
eVision.EasyDeepLearning.EDeepLearningDevice	GetActiveDevice ()
	Gets the active device. By default, it returns the first active device.
bool	HasEngineSetupData ()
	Whether the inference model has engine setup data.
void	ClearEngineSetupData ()
	Clears the engine data saved in the inference model. Use this method when your engine fails.
bool	IsTraining ()
	Indicates whether the object is currently training.
int	WaitForTrainingCompletion (int timeout)
	Waits until the training is complete or the timeout is expired. A call to this method will block the calling thread for the shortest time between the timeout and the time it takes for the training to complete.
int	WaitForTrainingCompletion ()
	Waits until the training is complete or the timeout is expired. A call to this method will block the calling thread for the shortest time between the timeout and the time it takes for the training to complete.
int	WaitForIterationCompletion ()
	Waits until an iteration is complete. A call to this method will block the calling thread until a training iteration in the training thread is finished. This method returns the number of trained iterations.
int	StopTraining (bool wait)
	Stops training and returns the last completed iteration. If the parameter 'wait' is set to true, the method will wait for the training thread to completely stop. Otherwise, the method will return immediately.
void	SetLearningRate (eVision.EasyDeepLearning.EDeepLearningLearningRateType type, float[] parameters)
	Specifies how the value of the learning rate and how it evolves during training.
void	InitializeInference (int width, int height, eVision.EImageType imageType, int numImages)
	Initializes the neural network for inference (i.e. making predictions with the neural network).
void	InitializeInference (eVision.EBaseROI img)
	Initializes the neural network for inference (i.e. making predictions with the neural network).
void	InitializeInference (ref eVision.EBaseROI[] imgList)
	Initializes the neural network for inference (i.e. making predictions with the neural network).
void	InitializeInference (ref eVision.EImageBW8[] imgList)
	Initializes the neural network for inference (i.e. making predictions with the neural network).
void	InitializeInference (ref eVision.EImageBW16[] imgList)
	Initializes the neural network for inference (i.e. making predictions with the neural network).
void	InitializeInference (ref eVision.EImageC24[] imgList)
	Initializes the neural network for inference (i.e. making predictions with the neural network).
string	GetLabel (uint index)
	Gets the label from its index. If the tool is not trained, the method may throw an exception.
virtual void	SetLabel (uint index, string label)
	Changes a label predicted by the tool.
void	SetLabelWeight (uint index, float weight)
	Sets the label weight. If the tool is not trained, the label weights are not defined. After training, they are set to the value of the training dataset label weights.
float	GetLabelWeight (uint index)
	Gets the label weight. If the tool is not trained, the label weights are not defined. After training, they are set to the value of the training dataset label weights.
void	SetLabelColor (uint index, eVision.ERGBColor c)
	Changes a label color.
eVision.ERGBColor	GetLabelColor (uint index)
	Label color. The label colors affect how the results will be displayed.
void	SetLabelOpacity (uint index, byte opacity)
	Changes a label opacity (between 0 for fully transparent and 255 for opaque color).
byte	GetLabelOpacity (uint index)
	Gets the label opacity (between 0 for fully transparent and 255 for opaque color). The label opacities affect with the label colors how the results will be displayed.
int	GetOptimalNumImagesForBatchSize (eVision.EBaseROI sampleImage)
	Optimal number of images to process together for inference given the batch size and EDeepLearningTool.GetNumPatchesForImage.
int	GetOptimalNumImagesForBatchSize (int imageWidth, int imageHeight)
	Optimal number of images to process together for inference given the batch size and EDeepLearningTool.GetNumPatchesForImage.
int	GetNumPatchesForImage (eVision.EBaseROI sampleImage)
	Number of patches that will be extracted from an input image to perform inference.
virtual int	GetNumPatchesForImage (int imageWidth, int imageHeight)
	Number of patches that will be extracted from an input image to perform inference.
virtual void	Clean ()
	Reset the engine and clean CPU and GPU memory allocated for running the model.

Static Public Member Functions
static eVision.EasyDeepLearning.EDeepLearningTool	Create (string path)
	Factory method to open a deep learning tool.
static eVision.EasyDeepLearning.EDeepLearningTool	Create (eVision.ESerializer serializer)
	Factory method to open a deep learning tool.
static eVision.EasyDeepLearning.EDeepLearningTool	Create (eVision.EasyDeepLearning.EDeepLearningToolType type)
	Factory method to open a deep learning tool.
static string	GetAvailableEngineName (int id)
	Gets the name of the engine given its id.
static eVision.EasyDeepLearning.EDeepLearningDevice[]	GetAvailableDevicesForEngine (string engine)
	Available devices for the specified engine.

Properties
eVision.EasyDeepLearning.EDeepLearningToolType	ToolType `[get]`
	Type of the deep learning tool.
bool	EnableGPU `[get, set]`
	Enable the use of a GPU for training and inference of the deep learning tool. If the active engine is not the default engine, EDeepLearningTool.EnableGPU will be false even if the active device has a GPU device type. Setting EDeepLearningTool.EnableGPU to any value when the active engine is not the default one will throw an exception.
uint[]	GPUIndexes `[get, set]`
	Indexes of the GPUs to use for computations.
uint	NumGPUs `[get]`
	Number of detected NVIDIA GPU.
string	Engine `[get, set]`
	Sets the engine with which to execute or train the neural network.
eVision.EasyDeepLearning.EDeepLearningInferencePrecision	InferencePrecision `[get, set]`
	Inference precision.
int	NumAvailableDevices `[get]`
	Available devices for the current EDeepLearningTool.Engine.
eVision.EasyDeepLearning.EDeepLearningDevice[]	AvailableDevices `[get]`
	Get all available device for the current engine.
int	NumActiveDevices `[get]`
	Gets the number of devices with which to run or train the model.
string	ActiveDeviceByName `[set]`
	Sets the device with which to run the model using its name.
int	ActiveDeviceById `[set]`
	Sets the device with which to run the model using its index in the list returned by EDeepLearningTool.AvailableDevices.
eVision.EasyDeepLearning.EDeepLearningDevice[]	ActiveDevices `[get, set]`
	Devices with which to run or train the model.
int[]	ActiveDevicesById `[set]`
	Sets the devices with which to run the model using their index in the list returned by EDeepLearningTool.AvailableDevices.
eVision.EasyDeepLearning.EDeepLearningExecutionSettings	ExecutionSettings `[get, set]`
	Execution settings of the tool (engine, device(s), inference precision, batch size settings).
eVision.EasyDeepLearning.EDeepLearningDevice	EngineSetupDataDevice `[get]`
	Device corresponding to the engine data contained in the tool.
eVision.EasyDeepLearning.EDeepLearningInferencePrecision	EngineSetupDataInferencePrecision `[get]`
	Inference precision corresponding to the engine data contained in the tool.
int	EngineSetupDataBatchSize `[get]`
	Batch size corresponding to the engine data contained in the tool.
bool	OptimizeBatchSize `[get, set]`
	Indicates whether to optimize the batch size (see EDeepLearningTool.BatchSize) to maximize the training and inference speed according to the engine, the device and the available memory.
int	BatchSizeForMaximumInferenceSpeed `[get]`
	Computes the batch size that will maximize the inference speed on NVIDIA GPU. For all other devices, it will return 1.
int	BatchSize `[get, set]`
	Batch size. The batch size is the number of images that are processed together during training and batch inference.
float	CurrentTrainingProgression `[get]`
	Current training progression as a percentage (between 0 and 1).
uint	CurrentTrainingFinishedIterations `[get]`
	Number of iterations that are already finished in the current training.
uint	CurrentTrainingNumIterations `[get]`
	Total number of training iterations that will be performed during the current training of the deep learning tool.
uint	NumTrainedIterations `[get]`
	Number of iterations that were performed to train this deep learning tool.
int	BestIteration `[get]`
	Iteration at which the minimum validation error was reached. After training, the tool is in the state it was at this best iteration.
ulong	ImageCacheSize `[get, set]`
	Size in byte of the image cache. The cache is used during training to store reformatted and normalized images. A correctly sized cache can reduce the hard drive accesses and the preprocessing time for each image.
string	Path `[get]`
	Path of the tool when loaded or saved using EDeepLearningTool.Save or EDeepLearningTool.Load.
string	InferenceModelPath `[get]`
	Path of the inference model when loaded or saved using EDeepLearningTool.SaveInferenceModel, EDeepLearningTool.LoadInferenceModel or EDeepLearningTool.SerializeInferenceModel.
string	TrainingModelPath `[get]`
	Path of the training model when loaded or saved using EDeepLearningTool.SaveTrainingModel, EDeepLearningTool.LoadTrainingModel or EDeepLearningTool.SerializeTrainingModel.
string	SettingsPath `[get]`
	Path of the settings when loaded or saved using EDeepLearningTool.SaveSettings, EDeepLearningTool.LoadSettings or EDeepLearningTool.SerializeSettings.
int	AutoSavePeriod `[get, set]`
	Auto-save period. The auto-save period is the number of training iterations between each automatic save of the model.
eVision.EasyDeepLearning.EDeepLearningLearningRateType	LearningRateType `[get]`
	Learning rate type.
float[]	LearningRateParameters `[get]`
	Parameters of the learning rate.
bool	EnableDeterministicTraining `[get, set]`
	Whether to use deterministic training algorithm that allows to repeatable training results. Default value: false.
int	DeterministicTrainingRandomSeed `[get, set]`
	Random seed to use when EDeepLearningTool.EnableDeterministicTraining is set to true. The default value is 42.
uint	NumLabels `[get]`
	Number of labels of this tool. If the tool is not trained, the method will throw an exception.
static int	NumAvailableEngines `[get]`
	Number of available engines.
static string[]	AvailableEngines `[get]`
	Available engines.

Detailed Description

EDeepLearningTool represents the common operations of deep learning tools.

The class is responsible for handling CPU/GPU settings and training. Computing the result for an image is called inference and is the responsibility of the actual deep learning tools (see EClassifier).

License(s): EasyClassify or EasySegment or EasyLocate or EasyClassifyInference or EasySegmentInference or EasyLocateInference

Member Function Documentation

◆ Clean()

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Clean ( )

virtual

Reset the engine and clean CPU and GPU memory allocated for running the model.

This method does not clean all GPU memory. For example, some program data (CUDA kernels) will be kept in GPU memory.

◆ ClearEngineSetupData()

void eVision.EasyDeepLearning.EDeepLearningTool.ClearEngineSetupData ( )

Clears the engine data saved in the inference model. Use this method when your engine fails.

◆ Clone()

eVision.EasyDeepLearning.EDeepLearningTool eVision.EasyDeepLearning.EDeepLearningTool.Clone ( )

Clones this tool.

◆ Create() [1/3]

eVision.EasyDeepLearning.EDeepLearningTool eVision.EasyDeepLearning.EDeepLearningTool.Create ( eVision.EasyDeepLearning.EDeepLearningToolType type )

static

Factory method to open a deep learning tool.

Parameters

type	Type of tool to create

◆ Create() [2/3]

eVision.EasyDeepLearning.EDeepLearningTool eVision.EasyDeepLearning.EDeepLearningTool.Create ( eVision.ESerializer serializer )

static

Factory method to open a deep learning tool.

Parameters

serializer A serializer created for reading.

◆ Create() [3/3]

eVision.EasyDeepLearning.EDeepLearningTool eVision.EasyDeepLearning.EDeepLearningTool.Create ( string path )

static

Factory method to open a deep learning tool.

Parameters

path	Path to a deep learning tool.

◆ GetActiveDevice() [1/2]

eVision.EasyDeepLearning.EDeepLearningDevice eVision.EasyDeepLearning.EDeepLearningTool.GetActiveDevice ( )

Gets the active device. By default, it returns the first active device.

◆ GetActiveDevice() [2/2]

eVision.EasyDeepLearning.EDeepLearningDevice eVision.EasyDeepLearning.EDeepLearningTool.GetActiveDevice ( int i )

Gets the active device. By default, it returns the first active device.

Parameters

i	Index of the active devices between 0 and EDeepLearningTool.NumActiveDevices.

◆ GetAvailableDevice()

eVision.EasyDeepLearning.EDeepLearningDevice eVision.EasyDeepLearning.EDeepLearningTool.GetAvailableDevice ( int i )

i-th EDeepLearningDevice of the current engine.

Parameters

i	Index of the device between 0 and EDeepLearningTool.NumAvailableDevices - 1.

◆ GetAvailableDeviceName()

string eVision.EasyDeepLearning.EDeepLearningTool.GetAvailableDeviceName ( int i )

Name of the i-th device of the current engine.

Parameters

i	Index of the device between 0 and EDeepLearningTool.NumAvailableDevices - 1.

◆ GetAvailableDevicesForEngine()

eVision.EasyDeepLearning.EDeepLearningDevice[] eVision.EasyDeepLearning.EDeepLearningTool.GetAvailableDevicesForEngine ( string engine )

static

Available devices for the specified engine.

Parameters

engine Engine for which to get the devices

◆ GetAvailableEngineName()

string eVision.EasyDeepLearning.EDeepLearningTool.GetAvailableEngineName ( int id )

static

Gets the name of the engine given its id.

Parameters

id	Index of the available engine between 0 and EDeepLearningTool.NumAvailableEngines - 1.

◆ GetLabel()

string eVision.EasyDeepLearning.EDeepLearningTool.GetLabel ( uint index )

Gets the label from its index. If the tool is not trained, the method may throw an exception.

Parameters

index Index of the label

◆ GetLabelColor()

eVision.ERGBColor eVision.EasyDeepLearning.EDeepLearningTool.GetLabelColor ( uint index )

Label color. The label colors affect how the results will be displayed.

The default label colors are the one specified in the dataset used for training.

Parameters

index Index of the label between 0 and EDeepLearningTool.NumLabels

◆ GetLabelOpacity()

byte eVision.EasyDeepLearning.EDeepLearningTool.GetLabelOpacity ( uint index )

Gets the label opacity (between 0 for fully transparent and 255 for opaque color). The label opacities affect with the label colors how the results will be displayed.

The default label opacities are the one specified in the dataset used for training.

Parameters

index Index of the label between 0 and EDeepLearningTool.NumLabels

◆ GetLabelWeight()

float eVision.EasyDeepLearning.EDeepLearningTool.GetLabelWeight ( uint index )

Gets the label weight. If the tool is not trained, the label weights are not defined. After training, they are set to the value of the training dataset label weights.

Parameters

index Index of the label

◆ GetNumPatchesForImage() [1/2]

int eVision.EasyDeepLearning.EDeepLearningTool.GetNumPatchesForImage ( eVision.EBaseROI sampleImage )

Number of patches that will be extracted from an input image to perform inference.

For EasyClassify and EasyLocate, this will always be equal to 1. For EasySegment, the number of patches will depend on the scale, patch size and sampling density parameters.

Parameters

sampleImage Image for which to get the number of patch

◆ GetNumPatchesForImage() [2/2]

virtual int eVision.EasyDeepLearning.EDeepLearningTool.GetNumPatchesForImage	(	int	imageWidth,
		int	imageHeight )

virtual

Number of patches that will be extracted from an input image to perform inference.

For EasyClassify and EasyLocate, this will always be equal to 1. For EasySegment, the number of patches will depend on the scale, patch size and sampling density parameters.

Parameters

imageWidth	Width of the image for which to get the number of patch
imageHeight	Height of the image for which to get the number of patch

Reimplemented in eVision.EasyDeepLearning.ESupervisedSegmenter, and eVision.EasyDeepLearning.EUnsupervisedSegmenter.

◆ GetOptimalNumImagesForBatchSize() [1/2]

int eVision.EasyDeepLearning.EDeepLearningTool.GetOptimalNumImagesForBatchSize ( eVision.EBaseROI sampleImage )

Optimal number of images to process together for inference given the batch size and EDeepLearningTool.GetNumPatchesForImage.

In practice, for EasyClassify and EasyLocate, this is the same as the batch size.

For EasySegment, the value will depend on the scale, patch size and sampling density parameters.

Parameters

sampleImage Image for which to get the optimal number of images

◆ GetOptimalNumImagesForBatchSize() [2/2]

int eVision.EasyDeepLearning.EDeepLearningTool.GetOptimalNumImagesForBatchSize	(	int	imageWidth,
		int	imageHeight )

Optimal number of images to process together for inference given the batch size and EDeepLearningTool.GetNumPatchesForImage.

In practice, for EasyClassify and EasyLocate, this is the same as the batch size.

For EasySegment, the value will depend on the scale, patch size and sampling density parameters.

Parameters

imageWidth	Width of the image for which to get the optimal number of images
imageHeight	Height of the image for which to get the optimal number of images

◆ HasEngineSetupData()

bool eVision.EasyDeepLearning.EDeepLearningTool.HasEngineSetupData ( )

Whether the inference model has engine setup data.

Engine setup data allows for quicker configuration of the engine when reusing the same device, inference precision and batch size.

Use EDeepLearningTool.EngineSetupDataDevice, EDeepLearningTool.EngineSetupDataInferencePrecision, and EDeepLearningTool.EngineSetupDataBatchSize to retrieve the device (and engine), inference precision, and batch size for which the tool has engine setup data.

The engine setup data is stored in the tool after the first inference or after calling EDeepLearningTool.InitializeInference if the engine supports it. At the moment, only the TensorRT engine supports setup data. If you save the tool, the engine setup data will be saved with the tool and can be reused when loading the tool again.

The engine setup data will only be replaced when using another device and/or inference precision and/or batch size with an engine that supports engine setup data. Otherwise, this setup data will remain (for example, using an engine that does not supports setup data will leave the pervious engine setup data intact).

You can clear the engine setup data by using EDeepLearningTool.ClearEngineSetupData.

◆ HasInferenceModel()

bool eVision.EasyDeepLearning.EDeepLearningTool.HasInferenceModel ( )

Whether the inference model is loaded or created when the tool is not yet trained.

◆ HasTrainingModel()

bool eVision.EasyDeepLearning.EDeepLearningTool.HasTrainingModel ( )

Whether the training model is loaded or created when the tool is not yet trained.

◆ InitializeInference() [1/6]

void eVision.EasyDeepLearning.EDeepLearningTool.InitializeInference ( eVision.EBaseROI img )

Initializes the neural network for inference (i.e. making predictions with the neural network).

You need to pass an image or a vector of images with the same characteristics (image resolution, image type and number of image) as the image or vector of images that you will use to make inference with this tool.

If EDeepLearningTool.OptimizeBatchSize is equal to true, the value of EDeepLearningTool.BatchSize will be recomputed.

Parameters

img	Image for which to prepare the tool for inference

◆ InitializeInference() [2/6]

void eVision.EasyDeepLearning.EDeepLearningTool.InitializeInference	(	int	width,
		int	height,
		eVision.EImageType	imageType,
		int	numImages )

Initializes the neural network for inference (i.e. making predictions with the neural network).

You need to pass an image or a vector of images with the same characteristics (image resolution, image type and number of image) as the image or vector of images that you will use to make inference with this tool.

If EDeepLearningTool.OptimizeBatchSize is equal to true, the value of EDeepLearningTool.BatchSize will be recomputed.

Parameters

width	Width of the images for which to prepare the tool for inference
height	Height of the images for which to prepare the tool for inference
imageType	Type of the images for which to prepare the tool for inference
numImages	Number of images for which to prepare the tool for inference

◆ InitializeInference() [3/6]

void eVision.EasyDeepLearning.EDeepLearningTool.InitializeInference ( ref eVision.EBaseROI[] imgList )

Initializes the neural network for inference (i.e. making predictions with the neural network).

You need to pass an image or a vector of images with the same characteristics (image resolution, image type and number of image) as the image or vector of images that you will use to make inference with this tool.

If EDeepLearningTool.OptimizeBatchSize is equal to true, the value of EDeepLearningTool.BatchSize will be recomputed.

Parameters

imgList Vector of images for which to prepare the tool for inference

◆ InitializeInference() [4/6]

void eVision.EasyDeepLearning.EDeepLearningTool.InitializeInference ( ref eVision.EImageBW16[] imgList )

Initializes the neural network for inference (i.e. making predictions with the neural network).

You need to pass an image or a vector of images with the same characteristics (image resolution, image type and number of image) as the image or vector of images that you will use to make inference with this tool.

If EDeepLearningTool.OptimizeBatchSize is equal to true, the value of EDeepLearningTool.BatchSize will be recomputed.

Parameters

imgList Vector of images for which to prepare the tool for inference

◆ InitializeInference() [5/6]

void eVision.EasyDeepLearning.EDeepLearningTool.InitializeInference ( ref eVision.EImageBW8[] imgList )

Initializes the neural network for inference (i.e. making predictions with the neural network).

You need to pass an image or a vector of images with the same characteristics (image resolution, image type and number of image) as the image or vector of images that you will use to make inference with this tool.

If EDeepLearningTool.OptimizeBatchSize is equal to true, the value of EDeepLearningTool.BatchSize will be recomputed.

Parameters

imgList Vector of images for which to prepare the tool for inference

◆ InitializeInference() [6/6]

void eVision.EasyDeepLearning.EDeepLearningTool.InitializeInference ( ref eVision.EImageC24[] imgList )

Initializes the neural network for inference (i.e. making predictions with the neural network).

You need to pass an image or a vector of images with the same characteristics (image resolution, image type and number of image) as the image or vector of images that you will use to make inference with this tool.

If EDeepLearningTool.OptimizeBatchSize is equal to true, the value of EDeepLearningTool.BatchSize will be recomputed.

Parameters

imgList Vector of images for which to prepare the tool for inference

◆ IsTrained()

bool eVision.EasyDeepLearning.EDeepLearningTool.IsTrained ( )

Tells whether the deep learning tool has been trained.

◆ IsTraining()

bool eVision.EasyDeepLearning.EDeepLearningTool.IsTraining ( )

Indicates whether the object is currently training.

◆ Load() [1/2]

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Load ( eVision.ESerializer serializer )

virtual

Loads a deep learning tool. The given ESerializer must have been created for reading.

Parameters

serializer Pointer to the ESerializer created for reading.

◆ Load() [2/2]

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Load ( string filePath )

virtual

Loads a deep learning tool. The given ESerializer must have been created for reading.

Parameters

filePath File path.

◆ LoadInferenceModel()

void eVision.EasyDeepLearning.EDeepLearningTool.LoadInferenceModel ( string filePath )

Loads the tool's inference model. The inference model is required to apply the tool on new images.

Parameters

filePath File path.

◆ LoadSettings()

void eVision.EasyDeepLearning.EDeepLearningTool.LoadSettings ( string filePath )

Loads the settings of a tool.

Parameters

filePath File path.

◆ LoadTrainingModel()

void eVision.EasyDeepLearning.EDeepLearningTool.LoadTrainingModel ( string filePath )

Loads the tool's training model. The training model is required to continue a training.

Parameters

filePath File path.

◆ Save() [1/4]

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Save ( eVision.ESerializer serializer )

virtual

Saves the settings, inference and training model of the deep learning tool.

The given ESerializer must have been created for writing.

Parameters

serializer Pointer to the ESerializer created for writing.

◆ Save() [2/4]

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Save	(	eVision.ESerializer	serializer,
		bool	includeTrainingModel )

virtual

Saves the settings, inference and training model of the deep learning tool.

The given ESerializer must have been created for writing.

Parameters

serializer	Pointer to the ESerializer created for writing.
includeTrainingModel	Whether to save the training model. The training model is required to continue training the tool.

◆ Save() [3/4]

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Save ( string filePath )

virtual

Saves the settings, inference and training model of the deep learning tool.

The given ESerializer must have been created for writing.

Parameters

filePath File path.

◆ Save() [4/4]

virtual void eVision.EasyDeepLearning.EDeepLearningTool.Save	(	string	filePath,
		bool	includeTrainingModel )

virtual

Saves the settings, inference and training model of the deep learning tool.

The given ESerializer must have been created for writing.

Parameters

filePath	File path.
includeTrainingModel	Whether to save the training model. The training model is required to continue training the tool.

◆ SaveInferenceModel()

void eVision.EasyDeepLearning.EDeepLearningTool.SaveInferenceModel ( string filePath )

Saves the tool's inference model to a file.

Parameters

filePath File path.

◆ SaveSettings()

void eVision.EasyDeepLearning.EDeepLearningTool.SaveSettings ( string filePath )

Saves the settings of the tool to a file.

Parameters

filePath File path.

◆ SaveTrainingModel()

void eVision.EasyDeepLearning.EDeepLearningTool.SaveTrainingModel ( string filePath )

Saves the tool's training model to a file.

Parameters

filePath File path.

◆ SerializeInferenceModel()

void eVision.EasyDeepLearning.EDeepLearningTool.SerializeInferenceModel ( eVision.ESerializer serializer )

Serializes the inference model of the tool.

Parameters

serializer Pointer to ESerializer

◆ SerializeSettings()

void eVision.EasyDeepLearning.EDeepLearningTool.SerializeSettings ( eVision.ESerializer serializer )

abstract

Serializes the tool settings.

Parameters

serializer Pointer to ESerializer

◆ SerializeTrainingModel()

void eVision.EasyDeepLearning.EDeepLearningTool.SerializeTrainingModel ( eVision.ESerializer serializer )

Serializes the training model of the tool.

The training model is necessary to continue training the tool.

Parameters

serializer Pointer to ESerializer

◆ SetActiveDevice()

void eVision.EasyDeepLearning.EDeepLearningTool.SetActiveDevice ( eVision.EasyDeepLearning.EDeepLearningDevice device )

Sets the device with which to run or train the model.

Parameters

device The device with which to run or train the model.

For CPU device, the number of thread is initialized from Easy.MaxNumberOfProcessingThreads.

Note that this can change EDeepLearningTool.InferencePrecision if the previously set value was incompatible with the device.

◆ SetLabel()

virtual void eVision.EasyDeepLearning.EDeepLearningTool.SetLabel	(	uint	index,
		string	label )

virtual

Changes a label predicted by the tool.

Parameters

index	Index of the label
label	New label

Be careful when changing one of the label of a trained tool. It can create incompatibilities with previous metrics or results computed with this tool or with the dataset used to trained the tool.

Some tools may not use this label API. For example EUnsupervisedSegmenter uses EUnsupervisedSegmenter.GoodLabel instead.

◆ SetLabelColor()

void eVision.EasyDeepLearning.EDeepLearningTool.SetLabelColor	(	uint	index,
		eVision.ERGBColor	c )

Changes a label color.

Parameters

index	Index of the label between 0 and EDeepLearningTool.NumLabels
c	New color for the specified label

◆ SetLabelOpacity()

void eVision.EasyDeepLearning.EDeepLearningTool.SetLabelOpacity	(	uint	index,
		byte	opacity )

Changes a label opacity (between 0 for fully transparent and 255 for opaque color).

Parameters

index	Index of the label between 0 and EDeepLearningTool.NumLabels
opacity	New opacity for the specified label

◆ SetLabelWeight()

void eVision.EasyDeepLearning.EDeepLearningTool.SetLabelWeight	(	uint	index,
		float	weight )

Sets the label weight. If the tool is not trained, the label weights are not defined. After training, they are set to the value of the training dataset label weights.

Parameters

index	Index of the label
weight	Weight

◆ SetLearningRate()

void eVision.EasyDeepLearning.EDeepLearningTool.SetLearningRate	(	eVision.EasyDeepLearning.EDeepLearningLearningRateType	type,
		float[]	parameters )

Specifies how the value of the learning rate and how it evolves during training.

◆ StopTraining()

int eVision.EasyDeepLearning.EDeepLearningTool.StopTraining ( bool wait )

Stops training and returns the last completed iteration. If the parameter 'wait' is set to true, the method will wait for the training thread to completely stop. Otherwise, the method will return immediately.

Parameters

wait	Whether to wait for the training to completely stop

◆ UnloadInferenceModel()

void eVision.EasyDeepLearning.EDeepLearningTool.UnloadInferenceModel ( )

Unloads the inference model.

◆ UnloadModels()

void eVision.EasyDeepLearning.EDeepLearningTool.UnloadModels ( )

Unloads the inference and training model.

◆ UnloadTrainingModel()

void eVision.EasyDeepLearning.EDeepLearningTool.UnloadTrainingModel ( )

Unloads the training model.

◆ WaitForIterationCompletion()

int eVision.EasyDeepLearning.EDeepLearningTool.WaitForIterationCompletion ( )

Waits until an iteration is complete. A call to this method will block the calling thread until a training iteration in the training thread is finished. This method returns the number of trained iterations.

◆ WaitForTrainingCompletion() [1/2]

int eVision.EasyDeepLearning.EDeepLearningTool.WaitForTrainingCompletion ( )

Waits until the training is complete or the timeout is expired. A call to this method will block the calling thread for the shortest time between the timeout and the time it takes for the training to complete.

A negative timeout means that the method will wait until the training is complete.

The default value is set to -1.

The method returns the number of trained iterations.

◆ WaitForTrainingCompletion() [2/2]

int eVision.EasyDeepLearning.EDeepLearningTool.WaitForTrainingCompletion ( int timeout )

Waits until the training is complete or the timeout is expired. A call to this method will block the calling thread for the shortest time between the timeout and the time it takes for the training to complete.

A negative timeout means that the method will wait until the training is complete.

The default value is set to -1.

The method returns the number of trained iterations.

Parameters

timeout Timeout in second

Property Documentation

◆ ActiveDeviceById

int eVision.EasyDeepLearning.EDeepLearningTool.ActiveDeviceById

set

Sets the device with which to run the model using its index in the list returned by EDeepLearningTool.AvailableDevices.

◆ ActiveDeviceByName

string eVision.EasyDeepLearning.EDeepLearningTool.ActiveDeviceByName

set

Sets the device with which to run the model using its name.

◆ ActiveDevices

eVision.EasyDeepLearning.EDeepLearningDevice [] eVision.EasyDeepLearning.EDeepLearningTool.ActiveDevices

getset

Devices with which to run or train the model.

Only multiple GPUs with the default engine is supported.

◆ ActiveDevicesById

int [] eVision.EasyDeepLearning.EDeepLearningTool.ActiveDevicesById

set

Sets the devices with which to run the model using their index in the list returned by EDeepLearningTool.AvailableDevices.

◆ AutoSavePeriod

int eVision.EasyDeepLearning.EDeepLearningTool.AutoSavePeriod

getset

Auto-save period. The auto-save period is the number of training iterations between each automatic save of the model.

In order for the model to be automatically saved during the training, the tool must have been loaded or saved to a file. The path where the tool will be saved is either indicated by EDeepLearningTool.Path or the three paths EDeepLearningTool.SettingsPath, EDeepLearningTool.InferenceModelPath, and EDeepLearningTool.TrainingModelPath. If one of these path is empty, the corresponding part of the tool will not be saved automatically during training.

By default, auto save is disabled and the period is equal to 0.

◆ AvailableDevices

eVision.EasyDeepLearning.EDeepLearningDevice [] eVision.EasyDeepLearning.EDeepLearningTool.AvailableDevices

get

Get all available device for the current engine.

◆ AvailableEngines

string [] eVision.EasyDeepLearning.EDeepLearningTool.AvailableEngines

staticget

Available engines.

◆ BatchSize

int eVision.EasyDeepLearning.EDeepLearningTool.BatchSize

getset

Batch size. The batch size is the number of images that are processed together during training and batch inference.

When using multi-GPUs processing (see EDeepLearningTool.GPUIndexes), the batch size is the number of images that each GPU will process at once.

A large batch size will increase the processing speed on GPU but also the memory requirements.

The batch size must be bigger or equal to 1 and it is commonly chosen to be a power of 2.

If EDeepLearningTool.OptimizeBatchSize is 'true', then the value of this property will be optimized automatically for training or inference according to the situation.

◆ BatchSizeForMaximumInferenceSpeed

int eVision.EasyDeepLearning.EDeepLearningTool.BatchSizeForMaximumInferenceSpeed

get

Computes the batch size that will maximize the inference speed on NVIDIA GPU. For all other devices, it will return 1.

This value is given as an indication and should not necessarily be used in practice.

You must choose a tradeoff between the overall inference speed (also called the throughput), which is limited by this value, and the time it takes to compute the result of a whole batch (also called the latency), which is minimized by making inference image per image (i.e. a batch size of 1). The tradeoff depends on your particular application.

Throw EError_DeepLearningToolNotTrained if the tool is not trained.

◆ BestIteration

int eVision.EasyDeepLearning.EDeepLearningTool.BestIteration

get

Iteration at which the minimum validation error was reached. After training, the tool is in the state it was at this best iteration.

◆ CurrentTrainingFinishedIterations

uint eVision.EasyDeepLearning.EDeepLearningTool.CurrentTrainingFinishedIterations

get

Number of iterations that are already finished in the current training.

◆ CurrentTrainingNumIterations

uint eVision.EasyDeepLearning.EDeepLearningTool.CurrentTrainingNumIterations

get

Total number of training iterations that will be performed during the current training of the deep learning tool.

After the end of the training, this number is the actual number of iterations performed during the training which can be lower than the number of iterations given to the training method (for example if the user called EDeepLearningTool.StopTraining).

◆ CurrentTrainingProgression

float eVision.EasyDeepLearning.EDeepLearningTool.CurrentTrainingProgression

get

Current training progression as a percentage (between 0 and 1).

◆ DeterministicTrainingRandomSeed

int eVision.EasyDeepLearning.EDeepLearningTool.DeterministicTrainingRandomSeed

getset

Random seed to use when EDeepLearningTool.EnableDeterministicTraining is set to true. The default value is 42.

When EDeepLearningTool.EnableDeterministicTraining, the random seed is generated using a random non-deterministic source.

◆ EnableDeterministicTraining

bool eVision.EasyDeepLearning.EDeepLearningTool.EnableDeterministicTraining

getset

Whether to use deterministic training algorithm that allows to repeatable training results. Default value: false.

Deterministic training is not guaranteed when:

using multi-core processing is enabled (see Easy.MaxNumberOfProcessingThreads) regardless of the value of EDeepLearningTool.EnableGPU).
performing several training on the same tool in succession.

◆ EnableGPU

bool eVision.EasyDeepLearning.EDeepLearningTool.EnableGPU

getset

Enable the use of a GPU for training and inference of the deep learning tool. If the active engine is not the default engine, EDeepLearningTool.EnableGPU will be false even if the active device has a GPU device type. Setting EDeepLearningTool.EnableGPU to any value when the active engine is not the default one will throw an exception.

Deprecated: Deprecated in version 23.4. Scheduled to be removed in version 27.6.

◆ Engine

string eVision.EasyDeepLearning.EDeepLearningTool.Engine

getset

Sets the engine with which to execute or train the neural network.

◆ EngineSetupDataBatchSize

int eVision.EasyDeepLearning.EDeepLearningTool.EngineSetupDataBatchSize

get

Batch size corresponding to the engine data contained in the tool.

◆ EngineSetupDataDevice

eVision.EasyDeepLearning.EDeepLearningDevice eVision.EasyDeepLearning.EDeepLearningTool.EngineSetupDataDevice

get

Device corresponding to the engine data contained in the tool.

◆ EngineSetupDataInferencePrecision

eVision.EasyDeepLearning.EDeepLearningInferencePrecision eVision.EasyDeepLearning.EDeepLearningTool.EngineSetupDataInferencePrecision

get

Inference precision corresponding to the engine data contained in the tool.

◆ ExecutionSettings

eVision.EasyDeepLearning.EDeepLearningExecutionSettings eVision.EasyDeepLearning.EDeepLearningTool.ExecutionSettings

getset

Execution settings of the tool (engine, device(s), inference precision, batch size settings).

◆ GPUIndexes

uint [] eVision.EasyDeepLearning.EDeepLearningTool.GPUIndexes

getset

Indexes of the GPUs to use for computations.

Using multiple GPUs is only possible when we can process multiple images at once, i.e. during training or batch inference.

By default, all the detected GPUs will be used.

The GPU are indexed from 0 to EDeepLearningTool.NumGPUs - 1.

Deprecated

Use EDeepLearningTool.GetActiveDevice or EDeepLearningTool.ActiveDevices.

Deprecated in version 23.4. Scheduled to be removed in version 27.6.

◆ ImageCacheSize

ulong eVision.EasyDeepLearning.EDeepLearningTool.ImageCacheSize

getset

Size in byte of the image cache. The cache is used during training to store reformatted and normalized images. A correctly sized cache can reduce the hard drive accesses and the preprocessing time for each image.

◆ InferenceModelPath

string eVision.EasyDeepLearning.EDeepLearningTool.InferenceModelPath

get

Path of the inference model when loaded or saved using EDeepLearningTool.SaveInferenceModel, EDeepLearningTool.LoadInferenceModel or EDeepLearningTool.SerializeInferenceModel.

◆ InferencePrecision

eVision.EasyDeepLearning.EDeepLearningInferencePrecision eVision.EasyDeepLearning.EDeepLearningTool.InferencePrecision

getset

Inference precision.

Use EDeepLearningDevice.IsPrecisionSupported to check whether a device support the given precision.

◆ LearningRateParameters

float [] eVision.EasyDeepLearning.EDeepLearningTool.LearningRateParameters

get

Parameters of the learning rate.

◆ LearningRateType

eVision.EasyDeepLearning.EDeepLearningLearningRateType eVision.EasyDeepLearning.EDeepLearningTool.LearningRateType

get

Learning rate type.

◆ NumActiveDevices

int eVision.EasyDeepLearning.EDeepLearningTool.NumActiveDevices

get

Gets the number of devices with which to run or train the model.

◆ NumAvailableDevices

int eVision.EasyDeepLearning.EDeepLearningTool.NumAvailableDevices

get

Available devices for the current EDeepLearningTool.Engine.

◆ NumAvailableEngines

int eVision.EasyDeepLearning.EDeepLearningTool.NumAvailableEngines

staticget

Number of available engines.

◆ NumGPUs

uint eVision.EasyDeepLearning.EDeepLearningTool.NumGPUs

get

Number of detected NVIDIA GPU.

Deprecated

Use EDeepLearningTool.NumAvailableDevices or parse the output of EDeepLearningTool.AvailableDevices to find the devices that are GPUs.

Deprecated in version 23.4. Scheduled to be removed in version 27.6.

◆ NumLabels

uint eVision.EasyDeepLearning.EDeepLearningTool.NumLabels

get

Number of labels of this tool. If the tool is not trained, the method will throw an exception.

Some type of tools such as EUnsupervisedSegmenter may have no label at all. For these tools, there is a separate API to control the predicted labels.

◆ NumTrainedIterations

uint eVision.EasyDeepLearning.EDeepLearningTool.NumTrainedIterations

get

Number of iterations that were performed to train this deep learning tool.

This number of iteration may result from the addition of the iterations performed in several calls to the training method.

An iteration can also be called an epoch.

◆ OptimizeBatchSize

bool eVision.EasyDeepLearning.EDeepLearningTool.OptimizeBatchSize

getset

Indicates whether to optimize the batch size (see EDeepLearningTool.BatchSize) to maximize the training and inference speed according to the engine, the device and the available memory.

Default value is true.

The batch size will be optimized at the start of training or during EDeepLearningTool.InitializeInference.

◆ Path

string eVision.EasyDeepLearning.EDeepLearningTool.Path

get

Path of the tool when loaded or saved using EDeepLearningTool.Save or EDeepLearningTool.Load.

◆ SettingsPath

string eVision.EasyDeepLearning.EDeepLearningTool.SettingsPath

get

Path of the settings when loaded or saved using EDeepLearningTool.SaveSettings, EDeepLearningTool.LoadSettings or EDeepLearningTool.SerializeSettings.

◆ ToolType

eVision.EasyDeepLearning.EDeepLearningToolType eVision.EasyDeepLearning.EDeepLearningTool.ToolType

getabstract

Type of the deep learning tool.

◆ TrainingModelPath

string eVision.EasyDeepLearning.EDeepLearningTool.TrainingModelPath

get

Path of the training model when loaded or saved using EDeepLearningTool.SaveTrainingModel, EDeepLearningTool.LoadTrainingModel or EDeepLearningTool.SerializeTrainingModel.

Public Member Functions

Static Public Member Functions

Properties

Detailed Description

Member Function Documentation

◆ Clean()

◆ ClearEngineSetupData()

◆ Clone()

◆ Create() [1/3]

◆ Create() [2/3]

◆ Create() [3/3]

◆ GetActiveDevice() [1/2]

◆ GetActiveDevice() [2/2]

◆ GetAvailableDevice()

◆ GetAvailableDeviceName()

◆ GetAvailableDevicesForEngine()

◆ GetAvailableEngineName()

◆ GetLabel()

◆ GetLabelColor()

◆ GetLabelOpacity()

◆ GetLabelWeight()

◆ GetNumPatchesForImage() [1/2]

◆ GetNumPatchesForImage() [2/2]

◆ GetOptimalNumImagesForBatchSize() [1/2]

◆ GetOptimalNumImagesForBatchSize() [2/2]

◆ HasEngineSetupData()

◆ HasInferenceModel()

◆ HasTrainingModel()

◆ InitializeInference() [1/6]

◆ InitializeInference() [2/6]

◆ InitializeInference() [3/6]

◆ InitializeInference() [4/6]

◆ InitializeInference() [5/6]

◆ InitializeInference() [6/6]

◆ IsTrained()

◆ IsTraining()

◆ Load() [1/2]

◆ Load() [2/2]

◆ LoadInferenceModel()

◆ LoadSettings()

◆ LoadTrainingModel()

◆ Save() [1/4]

◆ Save() [2/4]

◆ Save() [3/4]

◆ Save() [4/4]

◆ SaveInferenceModel()

◆ SaveSettings()

◆ SaveTrainingModel()

◆ SerializeInferenceModel()

◆ SerializeSettings()

◆ SerializeTrainingModel()

◆ SetActiveDevice()

◆ SetLabel()

◆ SetLabelColor()

◆ SetLabelOpacity()

◆ SetLabelWeight()

◆ SetLearningRate()

◆ StopTraining()

◆ UnloadInferenceModel()

◆ UnloadModels()

◆ UnloadTrainingModel()

◆ WaitForIterationCompletion()

◆ WaitForTrainingCompletion() [1/2]

◆ WaitForTrainingCompletion() [2/2]

Property Documentation

◆ ActiveDeviceById

◆ ActiveDeviceByName

◆ ActiveDevices

◆ ActiveDevicesById

◆ AutoSavePeriod

◆ AvailableDevices

◆ AvailableEngines

◆ BatchSize

◆ BatchSizeForMaximumInferenceSpeed

◆ BestIteration

◆ CurrentTrainingFinishedIterations

◆ CurrentTrainingNumIterations

◆ CurrentTrainingProgression

◆ DeterministicTrainingRandomSeed

◆ EnableDeterministicTraining