You are here: Start » FIL.NET » Invoke.TrainOcr_MLP

Invoke.TrainOcr_MLP

Trains an OCR multilayer perceptron classifier.

Namespace:Fil
Assembly:FilNet.dll

Syntax

C++
C#
 
public static void TrainOcr_MLP
(
	List<Fil.CharacterSample> inCharacterSamples,
	Fil.Size inNormalizationSize,
	Optional<List<int>> inHiddenLayerSizes,
	Optional<int> inRandomSeed,
	Fil.CharacterFeatures inCharacterFeatures,
	float inLearningRate,
	float inMomentum,
	int inIterationCount,
	Optional<Fil.Size> inCharacterSize,
	Fil.OcrModel outOcrModel,
	out float outTrainingAccuracy,
	Diagnostic<Fil.Profile> diagError,
	Diagnostic<List<Fil.Image>> diagNormalizedCharacters
)

Parameters

Name Type Range Default Description
inCharacterSamplesSystem.Collections.Generic.List<Fil.CharacterSample>Training font created from sample regions.
inNormalizationSizeFil.Size(Width: 16, Height: 16)The character size after normalization. Default value: (Width: 16, Height: 16).
inHiddenLayerSizesFtl.Optional<System.Collections.Generic.List<int>>Internal structure of neuron layers used in classifier. Default value: ftl::NIL.
inRandomSeedFtl.Optional<int><0, +INF>Random seed used by MLP classifier. Default value: ftl::NIL.
inCharacterFeaturesFil.CharacterFeatures(Pixels: True)Character features used to distinguish characters from each other. Default value: (Pixels: True).
inLearningRatefloat<0.01f, 1.0f>0.6fSuppression level of changes during learning process. Default value: 0.6f.
inMomentumfloat<0.0f, 1.0f>0.75fValue of classifier learning momentum. Default value: 0.75f.
inIterationCountint<1, +INF>100Learning iteration count. Default value: 100.
inCharacterSizeFtl.Optional<Fil.Size>Size of fixed width font. Default value: ftl::NIL.
outOcrModelFil.OcrModelTrained OcrMlpModel used to recognize characters.
outTrainingAccuracyfloatThe overall training score.
diagErrorFil.Diagnostic<Fil.Profile>Changes of mean error level progress during learning process.
diagNormalizedCharactersFil.Diagnostic<System.Collections.Generic.List<Fil.Image>>Images of normalized characters used to train classifier.

See also