/*
 * Copyright (C) 2005-2019 Centre National d'Etudes Spatiales (CNES)
 *
 * 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 otbTrainSVM_hxx
#define otbTrainSVM_hxx
#include "otbLearningApplicationBase.h"
#include "otbSVMMachineLearningModel.h"

namespace otb
{
namespace Wrapper
{

template <class TInputValue, class TOutputValue>
void LearningApplicationBase<TInputValue, TOutputValue>::InitSVMParams()
{
  AddChoice("classifier.svm", "SVM classifier (OpenCV)");
  SetParameterDescription("classifier.svm", "http://docs.opencv.org/modules/ml/doc/support_vector_machines.html");
  AddParameter(ParameterType_Choice, "classifier.svm.m", "SVM Model Type");
  SetParameterDescription("classifier.svm.m", "Type of SVM formulation.");
  if (this->m_RegressionFlag)
  {
    AddChoice("classifier.svm.m.epssvr", "Epsilon Support Vector Regression");
    AddChoice("classifier.svm.m.nusvr", "Nu Support Vector Regression");
    SetParameterString("classifier.svm.m", "epssvr");
  }
  else
  {
    AddChoice("classifier.svm.m.csvc", "C support vector classification");
    AddChoice("classifier.svm.m.nusvc", "Nu support vector classification");
    AddChoice("classifier.svm.m.oneclass", "Distribution estimation (One Class SVM)");
    SetParameterString("classifier.svm.m", "csvc");
  }
  AddParameter(ParameterType_Choice, "classifier.svm.k", "SVM Kernel Type");
  AddChoice("classifier.svm.k.linear", "Linear");

  AddChoice("classifier.svm.k.rbf", "Gaussian radial basis function");
  AddChoice("classifier.svm.k.poly", "Polynomial");
  AddChoice("classifier.svm.k.sigmoid", "Sigmoid");
  SetParameterString("classifier.svm.k", "linear");
  SetParameterDescription("classifier.svm.k", "SVM Kernel Type.");
  AddParameter(ParameterType_Float, "classifier.svm.c", "Cost parameter C");
  SetParameterFloat("classifier.svm.c", 1.0);
  SetParameterDescription("classifier.svm.c",
                          "SVM models have a cost parameter C (1 by default) to control the trade-off"
                          " between training errors and forcing rigid margins.");
  AddParameter(ParameterType_Float, "classifier.svm.nu", "Parameter nu of a SVM optimization problem (NU_SVC / ONE_CLASS)");
  SetParameterFloat("classifier.svm.nu", 0.0);
  SetParameterDescription("classifier.svm.nu", "Parameter nu of a SVM optimization problem.");
  if (this->m_RegressionFlag)
  {
    AddParameter(ParameterType_Float, "classifier.svm.p", "Parameter epsilon of a SVM optimization problem (EPS_SVR)");
    SetParameterFloat("classifier.svm.p", 1.0);
    SetParameterDescription("classifier.svm.p", "Parameter epsilon of a SVM optimization problem (EPS_SVR).");

    AddParameter(ParameterType_Choice, "classifier.svm.term", "Termination criteria");
    SetParameterDescription("classifier.svm.term", "Termination criteria for iterative algorithm");
    AddChoice("classifier.svm.term.iter", "Stops when maximum iteration is reached.");
    AddChoice("classifier.svm.term.eps", "Stops when accuracy is lower than epsilon.");
    AddChoice("classifier.svm.term.all", "Stops when either iteration or epsilon criteria is true");

    AddParameter(ParameterType_Float, "classifier.svm.iter", "Maximum iteration");
    SetParameterFloat("classifier.svm.iter", 1000);
    SetParameterDescription("classifier.svm.iter", "Maximum number of iterations (corresponds to the termination criteria 'iter').");

    AddParameter(ParameterType_Float, "classifier.svm.eps", "Epsilon accuracy threshold");
    SetParameterFloat("classifier.svm.eps", FLT_EPSILON);
    SetParameterDescription("classifier.svm.eps", "Epsilon accuracy (corresponds to the termination criteria 'eps').");
  }
  AddParameter(ParameterType_Float, "classifier.svm.coef0", "Parameter coef0 of a kernel function (POLY / SIGMOID)");
  SetParameterFloat("classifier.svm.coef0", 0.0);
  SetParameterDescription("classifier.svm.coef0", "Parameter coef0 of a kernel function (POLY / SIGMOID).");
  AddParameter(ParameterType_Float, "classifier.svm.gamma", "Parameter gamma of a kernel function (POLY / RBF / SIGMOID)");
  SetParameterFloat("classifier.svm.gamma", 1.0);
  SetParameterDescription("classifier.svm.gamma", "Parameter gamma of a kernel function (POLY / RBF / SIGMOID).");
  AddParameter(ParameterType_Float, "classifier.svm.degree", "Parameter degree of a kernel function (POLY)");
  SetParameterFloat("classifier.svm.degree", 1.0);
  SetParameterDescription("classifier.svm.degree", "Parameter degree of a kernel function (POLY).");
  AddParameter(ParameterType_Bool, "classifier.svm.opt", "Parameters optimization");
  SetParameterDescription("classifier.svm.opt",
                          "SVM parameters optimization flag.\n"
                          "-If set to True, then the optimal SVM parameters will be estimated. "
                          "Parameters are considered optimal by OpenCV when the cross-validation estimate of "
                          "the test set error is minimal. Finally, the SVM training process is computed "
                          "10 times with these optimal parameters over subsets corresponding to 1/10th of "
                          "the training samples using the k-fold cross-validation (with k = 10).\n-If set "
                          "to False, the SVM classification process will be computed once with the "
                          "currently set input SVM parameters over the training samples.\n-Thus, even "
                          "with identical input SVM parameters and a similar random seed, the output "
                          "SVM models will be different according to the method used (optimized or not) "
                          "because the samples are not identically processed within OpenCV.");
}

template <class TInputValue, class TOutputValue>
void LearningApplicationBase<TInputValue, TOutputValue>::TrainSVM(typename ListSampleType::Pointer trainingListSample,
                                                                  typename TargetListSampleType::Pointer trainingLabeledListSample, std::string modelPath)
{
  typedef otb::SVMMachineLearningModel<InputValueType, OutputValueType> SVMType;
  typename SVMType::Pointer SVMClassifier = SVMType::New();
  SVMClassifier->SetRegressionMode(this->m_RegressionFlag);
  SVMClassifier->SetInputListSample(trainingListSample);
  SVMClassifier->SetTargetListSample(trainingLabeledListSample);
  switch (GetParameterInt("classifier.svm.k"))
  {
  case 0: // LINEAR
    SVMClassifier->SetKernelType(CvSVM::LINEAR);
    std::cout << "CvSVM::LINEAR = " << CvSVM::LINEAR << std::endl;
    break;
  case 1: // RBF
    SVMClassifier->SetKernelType(CvSVM::RBF);
    std::cout << "CvSVM::RBF = " << CvSVM::RBF << std::endl;
    break;
  case 2: // POLY
    SVMClassifier->SetKernelType(CvSVM::POLY);
    std::cout << "CvSVM::POLY = " << CvSVM::POLY << std::endl;
    break;
  case 3: // SIGMOID
    SVMClassifier->SetKernelType(CvSVM::SIGMOID);
    std::cout << "CvSVM::SIGMOID = " << CvSVM::SIGMOID << std::endl;
    break;
  default: // DEFAULT = LINEAR
    SVMClassifier->SetKernelType(CvSVM::LINEAR);
    std::cout << "CvSVM::LINEAR = " << CvSVM::LINEAR << std::endl;
    break;
  }
  if (this->m_RegressionFlag)
  {
    switch (GetParameterInt("classifier.svm.m"))
    {
    case 0: // EPS_SVR
      SVMClassifier->SetSVMType(CvSVM::EPS_SVR);
      std::cout << "CvSVM::EPS_SVR = " << CvSVM::EPS_SVR << std::endl;
      break;
    case 1: // NU_SVR
      SVMClassifier->SetSVMType(CvSVM::NU_SVR);
      std::cout << "CvSVM::NU_SVR = " << CvSVM::NU_SVR << std::endl;
      break;
    default: // DEFAULT = EPS_SVR
      SVMClassifier->SetSVMType(CvSVM::EPS_SVR);
      std::cout << "CvSVM::EPS_SVR = " << CvSVM::EPS_SVR << std::endl;
      break;
    }
  }
  else
  {
    switch (GetParameterInt("classifier.svm.m"))
    {
    case 0: // C_SVC
      SVMClassifier->SetSVMType(CvSVM::C_SVC);
      std::cout << "CvSVM::C_SVC = " << CvSVM::C_SVC << std::endl;
      break;
    case 1: // NU_SVC
      SVMClassifier->SetSVMType(CvSVM::NU_SVC);
      std::cout << "CvSVM::NU_SVC = " << CvSVM::NU_SVC << std::endl;
      break;
    case 2: // ONE_CLASS
      SVMClassifier->SetSVMType(CvSVM::ONE_CLASS);
      std::cout << "CvSVM::ONE_CLASS = " << CvSVM::ONE_CLASS << std::endl;
      break;
    default: // DEFAULT = C_SVC
      SVMClassifier->SetSVMType(CvSVM::C_SVC);
      std::cout << "CvSVM::C_SVC = " << CvSVM::C_SVC << std::endl;
      break;
    }
  }
  SVMClassifier->SetC(GetParameterFloat("classifier.svm.c"));
  SVMClassifier->SetNu(GetParameterFloat("classifier.svm.nu"));
  if (this->m_RegressionFlag)
  {
    SVMClassifier->SetP(GetParameterFloat("classifier.svm.p"));
    switch (GetParameterInt("classifier.svm.term"))
    {
    case 0: // ITER
      SVMClassifier->SetTermCriteriaType(CV_TERMCRIT_ITER);
      break;
    case 1: // EPS
      SVMClassifier->SetTermCriteriaType(CV_TERMCRIT_EPS);
      break;
    case 2: // ITER+EPS
      SVMClassifier->SetTermCriteriaType(CV_TERMCRIT_ITER + CV_TERMCRIT_EPS);
      break;
    default:
      SVMClassifier->SetTermCriteriaType(CV_TERMCRIT_ITER);
      break;
    }
    SVMClassifier->SetMaxIter(GetParameterInt("classifier.svm.iter"));
    SVMClassifier->SetEpsilon(GetParameterFloat("classifier.svm.eps"));
  }
  SVMClassifier->SetCoef0(GetParameterFloat("classifier.svm.coef0"));
  SVMClassifier->SetGamma(GetParameterFloat("classifier.svm.gamma"));
  SVMClassifier->SetDegree(GetParameterFloat("classifier.svm.degree"));
  SVMClassifier->SetParameterOptimization(GetParameterInt("classifier.svm.opt"));
  SVMClassifier->Train();
  SVMClassifier->Save(modelPath);

  // Update the displayed parameters in the GUI after the training process, for further use of them
  SetParameterFloat("classifier.svm.c", static_cast<float>(SVMClassifier->GetOutputC()));
  SetParameterFloat("classifier.svm.nu", static_cast<float>(SVMClassifier->GetOutputNu()));
  if (this->m_RegressionFlag)
  {
    SetParameterFloat("classifier.svm.p", static_cast<float>(SVMClassifier->GetOutputP()));
  }
  SetParameterFloat("classifier.svm.coef0", static_cast<float>(SVMClassifier->GetOutputCoef0()));
  SetParameterFloat("classifier.svm.gamma", static_cast<float>(SVMClassifier->GetOutputGamma()));
  SetParameterFloat("classifier.svm.degree", static_cast<float>(SVMClassifier->GetOutputDegree()));
}

} // end namespace wrapper
} // end namespace otb

#endif