[localization] rig calibration now can use SIFT, CCTAG or SIFT_CCTAG

#49

src/software/RigCalibration/CMakeLists.txt

@@ -1,12 +1,10 @@
 IF(OpenMVG_BUILD_VOCTREE)
 
-    IF(OpenMVG_USE_CCTAG)
       ###
       # Calibrate a rig based on CCTag localizers
       ###
       ADD_EXECUTABLE(openMVG_main_rigCalibration main_rigCalibration.cpp)
       INSTALL(TARGETS openMVG_main_rigCalibration DESTINATION bin/)
       TARGET_LINK_LIBRARIES(openMVG_main_rigCalibration openMVG_rig openMVG_localization  openMVG_dataio openMVG_image openMVG_features vlsift ${CCTAG_LIBRARIES} ${BOOST_LIBRARIES}) #todo: delete vlsift
-    ENDIF(OpenMVG_USE_CCTAG)
 
 ENDIF(OpenMVG_BUILD_VOCTREE)

src/software/RigCalibration/main_rigCalibration.cpp

@@ -1,5 +1,7 @@
 #include <openMVG/localization/VoctreeLocalizer.hpp>
+#if HAVE_CCTAG
 #include <openMVG/localization/CCTagLocalizer.hpp>
+#endif
 #include <openMVG/rig/Rig.hpp>
 #include <openMVG/sfm/pipelines/localization/SfM_Localizer.hpp>
 #include <openMVG/image/image_io.hpp>
@@ -34,6 +36,53 @@ namespace po = boost::program_options;
 
 using namespace openMVG;
 
+enum DescriberType
+{
+  SIFT
+#if HAVE_CCTAG
+  ,CCTAG,
+  SIFT_CCTAG
+#endif
+};
+
+inline DescriberType stringToDescriberType(const std::string& describerType)
+{
+  if(describerType == "SIFT")
+    return DescriberType::SIFT;
+#if HAVE_CCTAG
+  if (describerType == "CCTAG")
+    return DescriberType::CCTAG;
+  if(describerType == "SIFT_CCTAG")
+    return DescriberType::SIFT_CCTAG;
+#endif
+  throw std::invalid_argument("Unsupported describer type "+describerType);
+}
+
+inline std::string describerTypeToString(DescriberType describerType)
+{
+  if(describerType == DescriberType::SIFT)
+    return "SIFT";
+  if (describerType == DescriberType::CCTAG)
+    return "CCTAG";
+  if(describerType == DescriberType::SIFT_CCTAG)
+    return "SIFT_CCTAG";
+  throw std::invalid_argument("Unrecognized DescriberType "+std::to_string(describerType));
+}
+
+std::ostream& operator<<(std::ostream& os, const DescriberType describerType)
+{
+  os << describerTypeToString(describerType);
+  return os;
+}
+
+std::istream& operator>>(std::istream &in, DescriberType &describerType)
+{
+  int i;
+  in >> i;
+  describerType = static_cast<DescriberType>(i);
+  return in;
+}
+
 std::string myToString(std::size_t i, std::size_t zeroPadding)
 {
   std::stringstream ss;
@@ -43,24 +92,45 @@ std::string myToString(std::size_t i, std::size_t zeroPadding)
 
 int main(int argc, char** argv)
 {
-  std::string sfmFilePath; //< the OpenMVG .json data file
-  std::string descriptorsFolder; //< the OpenMVG .json data file
-  std::string mediaFilepath; //< the media file to localize
-  localization::CCTagLocalizer::Parameters param = localization::CCTagLocalizer::Parameters();
+  // common parameters
+  std::string sfmFilePath;                //< the OpenMVG .json data file
+  std::string descriptorsFolder;          //< the OpenMVG .json data file
+  std::string mediaFilepath;              //< the media file to localize
+  std::string preset = features::describerPreset_enumToString(features::EDESCRIBER_PRESET::NORMAL_PRESET);               //< the preset for the feature extractor
+  std::string str_descriptorType = describerTypeToString(DescriberType::CCTAG);               //< the preset for the feature extractor
+  bool refineIntrinsics = false;
+  // parameters for voctree localizer
+  std::string vocTreeFilepath;            //< the vocabulary tree file
+  std::string weightsFilepath;            //< the vocabulary tree weights file
+  std::string algostring = "FirstBest";   //< the localization algorithm to use for the voctree localizer
+  size_t numResults = 10;                 //< number of documents to search when querying the voctree
+  // parameters for cctag localizer
+  size_t nNearestKeyFrames = 5;           //
+
 #if HAVE_ALEMBIC
   std::string exportFile = "trackedcameras.abc"; //!< the export file
 #endif
+
   std::size_t nCam = 3;
-  po::options_description desc("This program takes as input a media (image, image sequence, video) and a database (voctree, 3D structure data) \n"
-                               "and returns for each frame a pose estimation for the camera.");
+  po::options_description desc("This program is used to calibrate a camera rig composed of internally calibrated cameras");
   desc.add_options()
           ("help,h", "Print this message")
-          ("results,r", po::value<size_t>(&param._nNearestKeyFrames)->default_value(param._nNearestKeyFrames), "Number of images to retrieve in database")
           ("sfmdata,d", po::value<std::string>(&sfmFilePath)->required(), "The sfm_data.json kind of file generated by OpenMVG [it could be also a bundle.out to use an older version of OpenMVG]")
           ("siftPath,s", po::value<std::string>(&descriptorsFolder), "Folder containing the .desc. If not provided, it will be assumed to be parent(sfmdata)/matches [for the older version of openMVG it is the list.txt]")
           ("mediafile,m", po::value<std::string>(&mediaFilepath)->required(), "The folder path containing all the synchronised image subfolders assocated to each camera")
-          ("refineIntrinsics,", po::bool_switch(&param._refineIntrinsics), "Enable/Disable camera intrinsics refinement for each localized image")
-          ("nCameras", po::value<size_t>(&nCam)->default_value(nCam), "Enable/Disable camera intrinsics refinement for each localized image")
+          ("refineIntrinsics,", po::bool_switch(&refineIntrinsics), "Enable/Disable camera intrinsics refinement for each localized image")
+          ("nCameras", po::value<size_t>(&nCam)->default_value(nCam), "Number of cameras composing the rig")
+          ("preset,", po::value<std::string>(&preset)->default_value(preset), "Preset for the feature extractor when localizing a new image {LOW,NORMAL,HIGH,ULTRA}")
+          ("descriptors,", po::value<std::string>(&str_descriptorType)->default_value(str_descriptorType), "Type of descriptors to use")
+  // parameters for voctree localizer
+          ("voctree,t", po::value<std::string>(&vocTreeFilepath), "Filename for the vocabulary tree")
+          ("weights,w", po::value<std::string>(&weightsFilepath), "Filename for the vocabulary tree weights")
+          ("algorithm,", po::value<std::string>(&algostring)->default_value(algostring), "Algorithm type: FirstBest=0, BestResult=1, AllResults=2, Cluster=3" )
+          ("results,r", po::value<size_t>(&numResults)->default_value(numResults), "Number of images to retrieve in database")
+#if HAVE_CCTAG
+  // parameters for cctag localizer
+          ("nNearestKeyFrames", po::value<size_t>(&nNearestKeyFrames)->default_value(nNearestKeyFrames), "Number of images to retrieve in database")
+#endif
 #if HAVE_ALEMBIC
           ("export,e", po::value<std::string>(&exportFile)->default_value(exportFile), "Filename for the SfM_Data export file (where camera poses will be stored). Default : trackedcameras.json If Alambic is enable it will also export an .abc file of the scene with the same name")
 #endif
@@ -92,27 +162,87 @@ int main(int argc, char** argv)
     POPART_COUT("Usage:\n\n" << desc);
     return EXIT_FAILURE;
   }
+  // just debugging prints
   {
     POPART_COUT("Program called with the following parameters:");
     POPART_COUT("\tsfmdata: " << sfmFilePath);
     POPART_COUT("\tmediafile: " << mediaFilepath);
     POPART_COUT("\tsiftPath: " << descriptorsFolder);
-    POPART_COUT("\tresults: " << param._nNearestKeyFrames);
-    POPART_COUT("\trefineIntrinsics: " << param._refineIntrinsics);
+    POPART_COUT("\trefineIntrinsics: " << refineIntrinsics);
+    POPART_COUT("\tnCameras: " << nCam);
+    POPART_COUT("\tpreset: " << preset);
+    POPART_COUT("\tdescriptors: " << str_descriptorType);
+    if((DescriberType::SIFT==stringToDescriberType(str_descriptorType))
+#if HAVE_CCTAG
+            ||(DescriberType::SIFT_CCTAG==stringToDescriberType(str_descriptorType))
+#endif
+      )
+    {
+      // parameters for voctree localizer
+      POPART_COUT("\tvoctree: " << vocTreeFilepath);
+      POPART_COUT("\tweights: " << weightsFilepath);
+      POPART_COUT("\talgorithm: " << algostring);
+      POPART_COUT("\tresults: " << numResults);
+    }
+#if HAVE_CCTAG
+    else
+    {
+      POPART_COUT("\tnNearestKeyFrames: " << nNearestKeyFrames);
+    }
+#endif
+
+  }
+
+  localization::LocalizerParameters *param;
+
+  localization::ILocalizer *localizer;
+
+  DescriberType describer = stringToDescriberType(str_descriptorType);
+
+  // initialize the localizer according to the chosen type of describer
+  if((DescriberType::SIFT==describer)
+#if HAVE_CCTAG
+            ||(DescriberType::SIFT_CCTAG==describer)
+#endif
+      )
+  {
+    localizer = new localization::VoctreeLocalizer(sfmFilePath,
+                                           descriptorsFolder,
+                                           vocTreeFilepath,
+                                           weightsFilepath
+#if HAVE_CCTAG
+                                           , DescriberType::SIFT_CCTAG==describer
+#endif
+                                           );
+    param = new localization::VoctreeLocalizer::Parameters();
+    param->_featurePreset = features::describerPreset_stringToEnum(preset);
+    param->_refineIntrinsics = refineIntrinsics;
+    localization::VoctreeLocalizer::Parameters *casted = static_cast<localization::VoctreeLocalizer::Parameters *>(param);
+    casted->_algorithm = localization::VoctreeLocalizer::initFromString(algostring);;
+    casted->_numResults = numResults;
+  }
+#if HAVE_CCTAG
+  else
+  {
+    localizer = new localization::CCTagLocalizer(sfmFilePath, descriptorsFolder);
+    param = new localization::CCTagLocalizer::Parameters();
+    param->_featurePreset = features::describerPreset_stringToEnum(preset);
+    param->_refineIntrinsics = refineIntrinsics;
+    localization::CCTagLocalizer::Parameters *casted = static_cast<localization::CCTagLocalizer::Parameters *>(param);
+    casted->_nNearestKeyFrames = nNearestKeyFrames;
   }
+#endif 
 
-  // init the localizer
-  localization::CCTagLocalizer localizer(sfmFilePath, descriptorsFolder);
 
-  if(!localizer.isInit())
+  if(!localizer->isInit())
   {
     POPART_CERR("ERROR while initializing the localizer!");
     return EXIT_FAILURE;
   }
 
 #if HAVE_ALEMBIC
   dataio::AlembicExporter exporter(exportFile);
-  exporter.addPoints(localizer.getSfMData().GetLandmarks());
+  exporter.addPoints(localizer->getSfMData().GetLandmarks());
 #endif
 
   // Create a camera rig
@@ -161,8 +291,8 @@ int main(int argc, char** argv)
       std::vector<pair<IndexT, IndexT> > ids;
       auto detect_start = std::chrono::steady_clock::now();
       localization::LocalizationResult localizationResult;
-      localizer.localize(imageGrey,
-                         &param,
+      localizer->localize(imageGrey,
+                         param,
                          hasIntrinsics/*useInputIntrinsics*/,
                          queryIntrinsics,
                          localizationResult);