BING update2

modules/saliency/include/opencv2/saliency/saliencySpecializedClasses.hpp

@@ -144,8 +144,8 @@ class CV_EXPORTS_W ObjectnessBING : public Objectness
   ObjectnessBING();
   ~ObjectnessBING();
 
-  void read( const FileNode& fn );
-  void write( FileStorage& fs ) const;
+  void read();
+  void write() const;
 
   // Load trained model.
      int loadTrainedModel(std::string modelName = ""); // Return -1, 0, or 1 if partial, none, or all loaded
@@ -158,6 +158,8 @@ class CV_EXPORTS_W ObjectnessBING : public Objectness
      vector<float> getobjectnessValues();
 
      void setColorSpace(int clr = MAXBGR);
+     void set_modelName(string name);
+     void set_bbResDir(string dir);
 
      // Read matrix from binary file
      static bool matRead( const std::string& filename, Mat& M);
@@ -182,7 +184,7 @@ class CV_EXPORTS_W ObjectnessBING : public Objectness
 
      //TODO Probably remove this parameters
      //DataSetVOC &_voc; // The dataset for training, testing
-     std:: string _modelName, _trainDirSI, _bbResDir;
+     std:: string _modelName, _bbResDir;
 
      vecI _svmSzIdxs; // Indexes of active size. It's equal to _svmFilters.size() and _svmReW1f.rows
      Mat _svmFilter; // Filters learned at stage I, each is a _H by _W CV_32F matrix

modules/saliency/samples/computeSaliency.cpp

@@ -70,6 +70,7 @@ int main( int argc, char** argv )
 
   frame.copyTo( image );
 
+
   if( saliency_algorithm.find( "SPECTRAL_RESIDUAL" ) == 0 )
   {
     Mat saliencyMap;
@@ -89,11 +90,22 @@ int main( int argc, char** argv )
   else if( saliency_algorithm.find( "BING" ) == 0 )
   {
     vector<Vec4i> saliencyMap;
+    //Ptr<ObjectnessBING> bing = Saliency::create( "BING" );
+    //bing=static_cast<ObjectnessBING>(saliencyAlgorithm);
+
+    saliencyAlgorithm.staticCast<ObjectnessBING>()->set_modelName("home/puja/src/opencv_contrib/modules/saliency/src/ObjectnessTrainedModel/");
+    saliencyAlgorithm.staticCast<ObjectnessBING>()->set_bbResDir("/home/puja/src/opencv_contrib/modules/saliency/src/ObjectnessTrainedModel/");
+
+    //ObjectnessBING bing(saliencyAlgorithm);
+
+    //bing->set_modelName("home/puja/src/opencv_contrib/modules/saliency/src/ObjectnessTrainedModel/");
+    //bing->set_bbResDir("/home/puja/src/opencv_contrib/modules/saliency/src/ObjectnessTrainedModel/");
+
     if( saliencyAlgorithm->computeSaliency( image, saliencyMap ) )
     {
 
       std::cout << "-----------------OBJECTNESS-----------" << std::endl;
-      std::cout << "OBJ BB VECTOR SIZE" << saliencyMap.size() << std::endl;
+      std::cout << "OBJECTNESS BOUNDING BOX VECTOR SIZE" << saliencyMap.size() << std::endl;
       std::cout << " " << saliencyMap[0] << std::endl;
       std::cout << " " << saliencyMap[1] << std::endl;
       std::cout << " " << saliencyMap[2] << std::endl;

modules/saliency/src/ValStructVec.h

@@ -72,8 +72,8 @@ struct ValStructVec
 template<typename VT, typename ST>
 void ValStructVec<VT, ST>::append( const ValStructVec<VT, ST> &newVals, int startV )
 {
-  int sz = newVals.size();
-  for ( int i = 0; i < sz; i++ )
+  int newValsSize = newVals.size();
+  for ( int i = 0; i < newValsSize; i++ )
     pushBack( (float) ( ( i + 300 ) * startV )/*newVals(i)*/, newVals[i] );
 }
 

modules/saliency/src/objectnessBING.cpp

@@ -77,8 +77,19 @@ void ObjectnessBING::setColorSpace( int clr )
   _Clr = clr;
   _modelName = "/home/puja/src/opencv_contrib/modules/saliency/src/ObjectnessTrainedModel/"
       + string( format( "ObjNessB%gW%d%s", _base, _W, _clrName[_Clr] ).c_str() );
-  //  _trainDirSI = _voc.localDir + string(format("TrainS1B%gW%d%s/", _base, _W, _clrName[_Clr]).c_str());
-  _bbResDir = "/home/puja/src/opencv_contrib/modules/saliency/src/" + string(format("BBoxesB%gW%d%s/", _base, _W, _clrName[_Clr]).c_str());
+  _bbResDir = "/home/puja/src/opencv_contrib/modules/saliency/src/" + string( format( "BBoxesB%gW%d%s/", _base, _W, _clrName[_Clr] ).c_str() );
+}
+
+void ObjectnessBING::set_modelName( string name )
+{
+
+  _modelName = name + string( format( "ObjNessB%gW%d%s", _base, _W, _clrName[_Clr] ).c_str() );
+}
+
+void ObjectnessBING::set_bbResDir( string dir )
+{
+
+  _modelName = dir + string( format( "BBoxesB%gW%d%s/", _base, _W, _clrName[_Clr] ).c_str() );
 }
 
 int ObjectnessBING::loadTrainedModel( string modelName )  // Return -1, 0, or 1 if partial, none, or all loaded
@@ -388,17 +399,23 @@ void ObjectnessBING::getObjBndBoxesForSingleImage( Mat img, ValStructVec<float,
     printf( "Average time for predicting an image (%s) is %gs\n", _clrName[_Clr], tm.TimeInSeconds() );
   }
 
-  /*CmFile::MkDir(_bbResDir);
+  //Write on file the total number and the list of rectangles returned by objectess, one for each row.
+
+  CmFile::MkDir( _bbResDir );
   CStr fName = _bbResDir + "bb";
-  //vector<Vec4i> sortedBB = finalBoxes.getSortedStructVal();
-  FILE *f = fopen( _S( fName + ".txt" ), "w" );
-  fprintf( f, "%d\n", finalBoxes.size() );
-  for ( size_t k = 0; k < finalBoxes.size(); k++ ){
-    //fprintf(f, "%g, %s\n", finalBoxes(k), _S(strVec4i(finalBoxes[k])));
-    //fprintf(f, "%g, %s\n", finalBoxes(k), finalBoxes(k)[0],finalBoxes(k)[1],finalBoxes(k)[2],finalBoxes(k)[3]);
+  vector<Vec4i> sortedBB = finalBoxes.getSortedStructVal();
+  std::ofstream ofs;
+  ofs.open( _S( fName + ".txt" ), std::ofstream::out );
+  stringstream dim;
+  dim << sortedBB.size();
+  ofs << dim.str() << "\n";
+  for ( size_t k = 0; k < sortedBB.size(); k++ )
+  {
+    stringstream str;
+    str << sortedBB[k][0] << " " << sortedBB[k][1] << " " << sortedBB[k][2] << " " << sortedBB[k][3] << "\n";
+    ofs << str.str();
   }
-
-    fclose( f ); */
+  ofs.close();
 }
 
 struct MatchPathSeparator
@@ -454,12 +471,12 @@ vector<float> ObjectnessBING::getobjectnessValues()
   return objectnessValues;
 }
 
-void ObjectnessBING::read( const cv::FileNode& fn )
+void ObjectnessBING::read()
 {
 
 }
 
-void ObjectnessBING::write( cv::FileStorage& fs ) const
+void ObjectnessBING::write() const
 {
 
 }
@@ -472,15 +489,15 @@ bool ObjectnessBING::computeSaliencyImpl( const InputArray image, OutputArray ob
   // List of rectangles returned by objectess function in ascending order.
   // At the top there are the rectangles with lower values of ​​objectness, ie more
   // likely to have objects in them.
-  //vector<Vec4i>
-
   vector<Vec4i> sortedBB = finalBoxes.getSortedStructVal();
 
-  objBoundingBox.create( 1, sortedBB.size(), CV_MAKETYPE( CV_32S, CV_MAT_CN(objBoundingBox.type()) ) );
-  Mat obj = objBoundingBox.getMat();
+  //objBoundingBox.create( 1, sortedBB.size(), CV_MAKETYPE( CV_32S, CV_MAT_CN(objBoundingBox.type()) ) );
+  //Mat obj = objBoundingBox.getMat();
+
+  //for ( uint i = 0; i < sortedBB.size(); i++ )
+  //  obj.at<Vec4i>( i ) = sortedBB[i];
 
-  for ( uint i = 0; i < sortedBB.size(); i++ )
-    obj.at<Vec4i>( i ) = sortedBB[i];
+  Mat( sortedBB ).copyTo( objBoundingBox );
 
   // List of the rectangles' objectness value
   unsigned long int valIdxesSize = finalBoxes.getvalIdxes().size();