otbSOM.hxx

Go to the documentation of this file.

/*
 * Copyright (C) 2005-2022 Centre National d'Etudes Spatiales (CNES)
 * Copyright (C) 2007-2012 Institut Mines Telecom / Telecom Bretagne
 *
 * This file is part of Orfeo Toolbox
 *
 *     https://www.orfeo-toolbox.org/
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
 
#ifndef otbSOM_hxx
#define otbSOM_hxx
 
#include "otbSOM.h"
#include "itkImageRegionIteratorWithIndex.h"
#include "itkFixedArray.h"
#include "otbMacro.h"
#include "itkImageRegionIterator.h"
#include "itkMersenneTwisterRandomVariateGenerator.h"
 
namespace otb
{
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::SOM()
{
  this->SetNumberOfRequiredInputs(0);
  this->SetNumberOfRequiredOutputs(1);
 
  m_MapSize.Fill(10);
  m_NumberOfIterations = 10;
  m_BetaInit           = 1.0;
  m_BetaEnd            = 0.2;
  m_NeighborhoodSizeInit.Fill(3);
  m_MinWeight  = static_cast<ValueType>(0.0);
  m_MaxWeight  = static_cast<ValueType>(128.0);
  m_RandomInit = false;
  m_Seed       = 123574651;
}
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::~SOM()
{
}
 
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
void SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::UpdateMap(const NeuronType& sample, double beta, SizeType& radius)
{
  // output map pointer
  MapPointerType map = this->GetOutput(0);
 
  // typedefs
  typedef itk::ImageRegionIteratorWithIndex<MapType> IteratorType;
  typedef itk::FixedArray<double, MapType::ImageDimension> FixedArrayIndexType;
  typedef itk::Statistics::EuclideanDistanceMetric<FixedArrayIndexType> DistanceType;
  typename DistanceType::Pointer                                        distance = DistanceType::New();
 
  // winner index in the map
  IndexType  position = map->GetWinner(sample);
  NeuronType winner   = map->GetPixel(position);
 
  // Local neighborhood definition
  RegionType localRegion;
  IndexType  localIndex = position - radius;
  SizeType   localSize;
  for (unsigned int i = 0; i < MapType::ImageDimension; ++i)
  {
    localSize[i] = 2 * radius[i] + 1;
  }
  localRegion.SetIndex(localIndex);
  localRegion.SetSize(localSize);
  localRegion.Crop(map->GetLargestPossibleRegion());
  IteratorType it(map, localRegion);
 
  // Walk through the map, and evolve each neuron depending on its
  // distance to the winner.
  for (it.GoToBegin(); !it.IsAtEnd(); ++it)
  {
    NeuronType tempNeuron = it.Get();
    NeuronType newNeuron(tempNeuron.Size());
 
    FixedArrayIndexType positionFA, indexFA;
    positionFA[0] = position[0];
    positionFA[1] = position[1];
 
    indexFA[0] = it.GetIndex()[0];
    indexFA[1] = it.GetIndex()[1];
 
    double tempBeta = beta / (1 + distance->Evaluate(positionFA, indexFA));
 
    for (unsigned int i = 0; i < newNeuron.Size(); ++i)
    {
      newNeuron[i] = tempNeuron[i] + static_cast<typename NeuronType::ValueType>((sample[i] - tempNeuron[i]) * tempBeta);
    }
 
    it.Set(newNeuron);
  }
}
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
void SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::Step(unsigned int currentIteration)
{
  // Compute the new learning coefficient
  double newBeta = m_BetaFunctor(currentIteration, m_NumberOfIterations, m_BetaInit, m_BetaEnd);
 
  // Compute the new neighborhood size
  SizeType newSize = m_NeighborhoodSizeFunctor(currentIteration, m_NumberOfIterations, m_NeighborhoodSizeInit);
 
  // update the neurons map with each example of the training set.
  otbMsgDebugMacro(<< "Beta: " << newBeta << ", radius: " << newSize);
  for (typename ListSampleType::Iterator it = m_ListSample->Begin(); it != m_ListSample->End(); ++it)
  {
    UpdateMap(it.GetMeasurementVector(), newBeta, newSize);
  }
}
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
void SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::GenerateOutputInformation()
{
  Superclass::GenerateOutputInformation();
  this->GetOutput()->SetNumberOfComponentsPerPixel(m_ListSample->GetMeasurementVectorSize());
 
  IndexType index;
  index.Fill(0);
 
  RegionType region;
  region.SetIndex(index);
  region.SetSize(this->GetMapSize());
  this->GetOutput()->SetRegions(region);
}
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
void SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::AllocateOutputs()
{
  if (this->GetNumberOfOutputs() != 1)
    itkExceptionMacro(<< "Number of output image should be 1");
 
  // output neuron map fill
  MapPointerType map = this->GetOutput(0);
  map->Allocate();
}
 
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
void SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::GenerateData(void)
{
  this->AllocateOutputs();
  this->BeforeThreadedGenerateData();
 
  MapPointerType map = this->GetOutput(0);
 
  if (m_RandomInit)
  {
    typedef itk::Statistics::MersenneTwisterRandomVariateGenerator GeneratorType;
    typedef itk::ImageRegionIterator<MapType>                      IteratorType;
    GeneratorType::Pointer                                         generator = GeneratorType::New();
    generator->Initialize(m_Seed);
 
    NeuronType   neuronInit(m_ListSample->GetMeasurementVectorSize());
    IteratorType it(map, map->GetLargestPossibleRegion());
 
    for (it.GoToBegin(); !it.IsAtEnd(); ++it)
    {
      for (unsigned int i = 0; i < neuronInit.Size(); ++i)
      {
        neuronInit[i] =
            static_cast<typename NeuronType::ValueType>(generator->GetUniformVariate(static_cast<double>(m_MinWeight), static_cast<double>(m_MaxWeight)));
      }
      it.Set(neuronInit);
    }
  }
  else
  {
    NeuronType neuronInit(m_ListSample->GetMeasurementVectorSize());
    neuronInit.Fill(m_MaxWeight);
    map->FillBuffer(neuronInit);
  }
 
  // Step through the iterations
  for (unsigned int i = 0; i < m_NumberOfIterations; ++i)
  {
    // otbMsgDebugMacro(<<"Step "<<i+1<<" / "<<m_NumberOfIterations);
    std::cerr << "Step " << i + 1 << " / " << m_NumberOfIterations << "                         \r";
    Step(i);
  }
 
  this->AfterThreadedGenerateData();
}
template <class TListSample, class TMap, class TSOMLearningBehaviorFunctor, class TSOMNeighborhoodBehaviorFunctor>
void SOM<TListSample, TMap, TSOMLearningBehaviorFunctor, TSOMNeighborhoodBehaviorFunctor>::PrintSelf(std::ostream& os, itk::Indent indent) const
{
  Superclass::PrintSelf(os, indent);
}
 
} // end namespace otb
 
#endif