Skip to content
Snippets Groups Projects
Commit 82e47317 authored by Cédric Traizet's avatar Cédric Traizet
Browse files

some cmments and documentation added

parent 0c0351e2
No related branches found
No related tags found
1 merge request!4Dimensionality reduction algorithms
Hide whitespace changes
Inline Side-by-side
include/AutoencoderModel.h
+ 7
− 6
View file @ 82e47317
......@@ -27,7 +27,7 @@ public:
itkNewMacro(Self);
itkTypeMacro(AutoencoderModel, DimensionalityReductionModel);
unsigned int GetDimension() {return m_NumberOfHiddenNeurons;};
unsigned int GetDimension() {return m_NumberOfHiddenNeurons;}; // Override the Dimensionality Reduction model method, it is used in the dimensionality reduction filter to set the output image size
itkGetMacro(NumberOfHiddenNeurons,unsigned int);
itkSetMacro(NumberOfHiddenNeurons,unsigned int);
......@@ -47,9 +47,7 @@ public:
void Load(const std::string & filename, const std::string & name="") ITK_OVERRIDE;
void Train() ITK_OVERRIDE;
//void Dimensionality_reduction() {}; // Dimensionality reduction is done by DoPredict
protected:
AutoencoderModel();
~AutoencoderModel() ITK_OVERRIDE;
......@@ -59,11 +57,14 @@ protected:
private:
/** Network attributes */
AutoencoderType m_net;
unsigned int m_NumberOfHiddenNeurons;
/** Training parameters */
unsigned int m_NumberOfIterations;
double m_Regularization;
double m_Noise;
double m_Regularization; // L2 Regularization parameter
double m_Noise; // probability for an input to be set to 0 (denosing autoencoder)
};
} // end namespace otb
......
include/AutoencoderModel.txx
+ 4
− 4
View file @ 82e47317
......@@ -6,12 +6,13 @@
#include <shark/Data/Dataset.h>
#include "itkMacro.h"
#include "otbSharkUtils.h"
//include train function
#include <shark/ObjectiveFunctions/ErrorFunction.h>
#include <shark/Algorithms/GradientDescent/Rprop.h>// the RProp optimization algorithm
#include <shark/ObjectiveFunctions/Loss/SquaredLoss.h> // squared loss used for regression
#include <shark/ObjectiveFunctions/Regularizer.h> //L2 regulariziation
#include <shark/Models/ImpulseNoiseModel.h>//noise source to corrupt the inputs
#include <shark/Models/ImpulseNoiseModel.h> //noise source to corrupt the inputs
#include <shark/Models/ConcatenatedModel.h>//to concatenate the noise with the model
namespace otb
......@@ -43,7 +44,7 @@ void AutoencoderModel<TInputValue,AutoencoderType>::Train()
std::size_t inputs = dataDimension(inputSamples);
m_net.setStructure(inputs, m_NumberOfHiddenNeurons);
initRandomUniform(m_net,-0.1*std::sqrt(1.0/inputs),0.1*std::sqrt(1.0/inputs));
shark::ImpulseNoiseModel noise(m_Noise,0.0);//set an input pixel with probability p to 0
shark::ImpulseNoiseModel noise(m_Noise,0.0); //set an input pixel with probability m_Noise to 0
shark::ConcatenatedModel<shark::RealVector,shark::RealVector> model = noise>> m_net;
shark::LabeledData<shark::RealVector,shark::RealVector> trainSet(inputSamples,inputSamples);//labels identical to inputs
......@@ -92,7 +93,7 @@ template <class TInputValue, class AutoencoderType>
void AutoencoderModel<TInputValue,AutoencoderType>::Save(const std::string & filename, const std::string & name)
{
std::ofstream ofs(filename);
ofs << m_net.name() << std::endl; //first line
ofs << m_net.name() << std::endl; // the first line of the model file contains a key
boost::archive::polymorphic_text_oarchive oa(ofs);
m_net.write(oa);
ofs.close();
......@@ -163,6 +164,5 @@ void AutoencoderModel<TInputValue,AutoencoderType>
}
}
} // namespace otb
#endif
include/DimensionalityReductionModel.h
+ 23
− 41
View file @ 82e47317
......@@ -25,42 +25,32 @@
namespace otb
{
/** \class MachineLearningModel
* \brief MachineLearningModel is the base class for all classifier objects (SVM, KNN,
* Random Forests, Artificial Neural Network, ...) implemented in the supervised classification framework of the OTB.
/** \class DimensionalityReductionModel
* \brief DimensionalityReductionModel is the base class for all dimensionality Reduction objects (PCA, autoencoders and SOM) implemented in the dimensionality Reduction framework of the OTB.
*
* MachineLearningModel is an abstract object that specifies behavior and
* interface of supervised classifiers (SVM, KNN, Random Forests, Artificial
* Neural Network, ...) in the generic supervised classification framework of the OTB.
* The main generic virtual methods specifically implemented in each classifier
* derived from the MachineLearningModel class are two learning-related methods:
* Train() and Save(), and three classification-related methods: Load(),
* DimensionalityReductionModel is an abstract object that specifies behavior and
* interface of dimensionality reduction algorithms (PCA, autoencoders and SOM) in the generic dimensionality Reduction framework of the OTB.
* The main generic virtual methods specifically implemented in each model
* derived from the DimensionalityReductionModel class are two learning-related methods:
* Train() and Save(), and three dimensionality reduction related methods: Load(),
* DoPredict() and optionnaly DoPredictBatch().
*
* Thus, each classifier derived from the MachineLearningModel class
* computes its corresponding model with Train() and exports it with
* the help of the Save() method.
*
* It is also possible to classify any input sample composed of several
* It is also possible to reduce the dimensionality of any input sample composed of several
* features (or any number of bands in the case of a pixel extracted
* from a multi-band image) with the help of the Predict() method which
* needs a previous loading of the classification model with the Load() method.
*
* \sa MachineLearningModelFactory
* \sa LibSVMMachineLearningModel
* \sa SVMMachineLearningModel
* \sa BoostMachineLearningModel
* \sa KNearestNeighborsMachineLearningModel
* \sa DecisionTreeMachineLearningModel
* \sa RandomForestsMachineLearningModel
* \sa GradientBoostedTreeMachineLearningModel
* \sa NormalBayesMachineLearningModel
* \sa NeuralNetworkMachineLearningModel
* \sa SharkRandomForestsMachineLearningModel
* \sa ImageClassificationFilter
* \sa DimensionalityReductionModelFactory
* \sa SOMModel
* \sa PCAModel
* \sa AutoencderModel
* \sa ImageDimensionalityReductionFilter
*
*
* \ingroup OTBSupervised
* \ingroup cbDimensionalityReduction
*/
template <class TInputValue, class TTargetValue>
class ITK_EXPORT DimensionalityReductionModel
......@@ -121,7 +111,7 @@ public:
/** Get the size of the output after dimensionality reduction */
virtual unsigned int GetDimension() = 0;
/**\name Classification model file manipulation */
/**\name Dimensionality Reduction model file manipulation */
//@{
/** Save the model to file */
virtual void Save(const std::string & filename, const std::string & name="") = 0;
......@@ -132,10 +122,10 @@ public:
/**\name Classification model file compatibility tests */
//@{
/** Is the input model file readable and compatible with the corresponding classifier ? */
/** Is the input model file readable and compatible with the corresponding model ? */
virtual bool CanReadFile(const std::string &) = 0;
/** Is the input model file writable and compatible with the corresponding classifier ? */
/** Is the input model file writable and compatible with the corresponding model ? */
virtual bool CanWriteFile(const std::string &) = 0;
//@}
......@@ -165,19 +155,13 @@ protected:
/** Is DoPredictBatch multi-threaded ? */
bool m_IsDoPredictBatchMultiThreaded;
private:
/** Actual implementation of BatchPredicition
* Default implementation will call DoPredict iteratively
* \param input The input batch
* \param startIndex Index of the first sample to predict
* \param size Number of samples to predict
* \param target Pointer to the list of produced labels
* \param quality Pointer to the list of produced confidence
* values, or NULL
* \param startIndex Index of the first sample to reduce
* \param size Number of samples to reduce
* \param target Pointer to the list of reduced samples
*
* Override me if internal implementation allows for batch
* prediction.
......@@ -187,11 +171,9 @@ private:
*/
virtual void DoPredictBatch(const InputListSampleType * input, const unsigned int & startIndex, const unsigned int & size, TargetListSampleType * target) const;
/** Actual implementation of single sample prediction
* \param input sample to predict
* \param quality Pointer to a variable to store confidence value,
* or NULL
* \return The predicted label
/** Actual implementation of single sample reduction
* \param input sample to reduce
* \return The reduced sample
*/
virtual TargetSampleType DoPredict(const InputSampleType& input) const = 0;
......
include/ImageDimensionalityReductionFilter.txx
+ 1
− 6
View file @ 82e47317
......@@ -163,14 +163,10 @@ ImageDimensionalityReductionFilter<TInputImage, TOutputImage, TMaskImage>
InputIteratorType inIt(inputPtr, outputRegionForThread);
OutputIteratorType outIt(outputPtr, outputRegionForThread);
// typedef typename ModelType::InputValueType InputValueType;
typedef typename ModelType::InputSampleType InputSampleType;
typedef typename ModelType::InputListSampleType InputListSampleType;
typedef typename ModelType::TargetValueType TargetValueType;
// typedef typename ModelType::TargetSampleType TargetSampleType;
typedef typename ModelType::TargetListSampleType TargetListSampleType;
// typedef typename ModelType::ConfidenceValueType ConfidenceValueType;
// typedef typename ModelType::ConfidenceSampleType ConfidenceSampleType;
typename InputListSampleType::Pointer samples = InputListSampleType::New();
unsigned int num_features = inputPtr->GetNumberOfComponentsPerPixel();
......@@ -193,11 +189,10 @@ ImageDimensionalityReductionFilter<TInputImage, TOutputImage, TMaskImage>
typename TargetListSampleType::Pointer labels;
// This call is threadsafe
//labels = m_Model->PredictBatch(samples,confidences);
labels = m_Model->PredictBatch(samples);
// Set the output values
typename TargetListSampleType::ConstIterator labIt = labels->Begin();
for (outIt.GoToBegin(); !outIt.IsAtEnd(); ++outIt)
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment