several enhancements

modules/datasetstools/doc/datasetstools.rst

@@ -75,9 +75,11 @@ _`"ChaLearn Looking at People"`: http://gesture.chalearn.org/
 
  1. Follow instruction from site above, download files for dataset "Track 3: Gesture Recognition": Train1.zip-Train5.zip, Validation1.zip-Validation3.zip (Register on site: www.codalab.org and accept the terms and conditions of competition: https://www.codalab.org/competitions/991#learn_the_details There are three mirrors for downloading dataset files. When I downloaded data only mirror: "Universitat Oberta de Catalunya" works).
 
- 2. Unpack train archives Train1.zip-Train5.zip to one folder (currently loading validation files wasn't implemented)
+ 2. Unpack train archives Train1.zip-Train5.zip to folder Train/, validation archives Validation1.zip-Validation3.zip to folder Validation/
 
- 3. To load data run: ./opencv/build/bin/example_datasetstools_gr_chalearn -p=/home/user/path_to_unpacked_folder/
+ 3. Unpack all archives in Train/ & Validation/ in the folders with the same names, for example: Sample0001.zip to Sample0001/
+
+ 4. To load data run: ./opencv/build/bin/example_datasetstools_gr_chalearn -p=/home/user/path_to_unpacked_folders/
 
 GR_skig
 =======
@@ -245,7 +247,7 @@ OR_sun
 
 Implements loading dataset:
 
-_`"SUN Database"`: http://sun.cs.princeton.edu/
+_`"SUN Database"`: http://sundatabase.mit.edu/
 
 Currently implemented loading "Scene Recognition Benchmark. SUN397". Planned to implement also "Object Detection Benchmark. SUN2012". 
 

modules/datasetstools/include/opencv2/datasetstools/dataset.hpp

@@ -66,10 +66,12 @@ public:
 
     std::vector< Ptr<Object> >& getTrain() { return train; }
     std::vector< Ptr<Object> >& getTest() { return test; }
+    std::vector< Ptr<Object> >& getValidation() { return validation; }
 
 protected:
     std::vector< Ptr<Object> > train;
     std::vector< Ptr<Object> > test;
+    std::vector< Ptr<Object> > validation;
 };
 
 }

modules/datasetstools/include/opencv2/datasetstools/ir_robot.hpp

@@ -59,10 +59,15 @@ namespace datasetstools
 // 0.0000e+00   2.8285e+03   6.1618e+02
 // 0.0000e+00   0.0000e+00   1.0000e+00
 
+struct cameraPos
+{
+    std::vector<std::string> images;
+};
+
 struct IR_robotObj : public Object
 {
     std::string name;
-    std::vector<std::string> images; // TODO: implement more complex structure
+    std::vector<cameraPos> pos;
 };
 
 class CV_EXPORTS IR_robot : public Dataset

modules/datasetstools/include/opencv2/datasetstools/msm_epfl.hpp

@@ -54,10 +54,21 @@ namespace cv
 namespace datasetstools
 {
 
+struct cameraParam
+{
+    Matx33d mat1;
+    double mat2[3];
+    Matx33d mat3;
+    double mat4[3];
+    int imageWidth, imageHeight;
+};
+
 struct MSM_epflObj : public Object
 {
     std::string imageName;
-    std::vector<double> bounding, camera, p; // TODO: implement better structures
+    Matx23d bounding;
+    Matx34d p;
+    cameraParam camera;
 };
 
 class CV_EXPORTS MSM_epfl : public Dataset

modules/datasetstools/samples/gr_chalearn.cpp

@@ -71,7 +71,8 @@ int main(int argc, char *argv[])
     // ***************
     // dataset contains information for each sample.
     // For example, let output dataset size and first element.
-    printf("dataset size: %u\n", (unsigned int)dataset->getTrain().size());
+    printf("train size: %u\n", (unsigned int)dataset->getTrain().size());
+    printf("validation size: %u\n", (unsigned int)dataset->getValidation().size());
     GR_chalearnObj *example = static_cast<GR_chalearnObj *>(dataset->getTrain()[0].get());
     printf("first dataset sample:\n%s\n", example->name.c_str());
     printf("color video:\n%s\n", example->nameColor.c_str());

modules/datasetstools/samples/ir_robot.cpp

@@ -72,12 +72,18 @@ int main(int argc, char *argv[])
     // dataset contains object with name and its images.
     // For example, let output last element and dataset size.
     IR_robotObj *example = static_cast<IR_robotObj *>(dataset->getTrain().back().get());
-    printf("last dataset object:\n%s\n", example->name.c_str());
+    printf("last dataset object:\n");
+    printf("name: %s\n", example->name.c_str());
+    printf("number postitions: %u\n", (unsigned int)example->pos.size());
     string currPath(path + example->name + "/");
-    for (vector<string>::iterator it=example->images.begin(); it!=example->images.end(); ++it)
+
+    for (vector<cameraPos>::iterator itP=example->pos.begin(); itP!=example->pos.end(); ++itP)
+    {
+        for (vector<string>::iterator it=itP->images.begin(); it!=itP->images.end(); ++it)
         {
             printf("%s\n", (currPath+(*it)).c_str());
         }
+    }
     printf("dataset size: %u\n", (unsigned int)dataset->getTrain().size());
 
     return 0;

modules/datasetstools/samples/msm_epfl.cpp

@@ -75,26 +75,60 @@ int main(int argc, char *argv[])
     MSM_epflObj *example = static_cast<MSM_epflObj *>(dataset->getTrain()[0].get());
     printf("first image:\nname: %s\n", example->imageName.c_str());
 
-    printf("bounding:\n");
+    printf("\nbounding:\n");
-    for (vector<double>::iterator it=example->bounding.begin(); it!=example->bounding.end(); ++it)
+    for (int i=0; i<2; ++i)
     {
-        printf("%f ", *it);
+        for (int j=0; j<3; ++j)
+        {
+            printf("%f ", example->bounding(i, j));
         }
         printf("\n");
+    }
 
-    printf("camera:\n");
+    printf("\ncamera:\n");
-    for (vector<double>::iterator it=example->camera.begin(); it!=example->camera.end(); ++it)
+    for (int i=0; i<3; ++i)
+    {
+        for (int j=0; j<3; ++j)
         {
-        printf("%f ", *it);
+            printf("%f ", example->camera.mat1(i, j));
+        }
+        printf("\n");
     }
     printf("\n");
 
-    printf("P:\n");
+    for (int i=0; i<3; ++i)
-    for (vector<double>::iterator it=example->p.begin(); it!=example->p.end(); ++it)
     {
-        printf("%f ", *it);
+        printf("%f ", example->camera.mat2[i]);
+    }
+    printf("\n\n");
+
+    for (int i=0; i<3; ++i)
+    {
+        for (int j=0; j<3; ++j)
+        {
+            printf("%f ", example->camera.mat3(i, j));
+        }
+        printf("\n");
     }
     printf("\n");
 
+    for (int i=0; i<3; ++i)
+    {
+        printf("%f ", example->camera.mat4[i]);
+    }
+    printf("\n\n");
+
+    printf("image width: %u, height: %u\n", example->camera.imageWidth, example->camera.imageHeight);
+
+    printf("\nP:\n");
+    for (int i=0; i<3; ++i)
+    {
+        for (int j=0; j<4; ++j)
+        {
+            printf("%f ", example->p(i, j));
+        }
+        printf("\n");
+    }
+
     return 0;
 }

modules/datasetstools/samples/slam_tumindoor.cpp

@@ -86,7 +86,7 @@ int main(int argc, char *argv[])
     }
     printf("file name: %s\n", (imagePath + example->name).c_str());
 
-    printf("transformation matrix:\n");
+    printf("\ntransformation matrix:\n");
     for (unsigned int i=0; i<4; ++i)
     {
         for (unsigned int j=0; j<4; ++j)

modules/datasetstools/samples/tr_chars.cpp

@@ -85,15 +85,20 @@ int main(int argc, char *argv[])
 
     vector< Ptr<Object> > &currTrain = dataset.back()->getTrain();
     vector< Ptr<Object> > &currTest = dataset.back()->getTest();
+    vector< Ptr<Object> > &currValidation = dataset.back()->getValidation();
     printf("train size: %u\n", (unsigned int)currTrain.size());
     printf("test size: %u\n", (unsigned int)currTest.size());
+    printf("validation size: %u\n", (unsigned int)currValidation.size());
 
-    TR_charsObj *example1 = static_cast<TR_charsObj *>(currTrain[0].get());
+    TR_charsObj *exampleTrain = static_cast<TR_charsObj *>(currTrain[0].get());
-    TR_charsObj *example2 = static_cast<TR_charsObj *>(currTest[0].get());
+    TR_charsObj *exampleTest = static_cast<TR_charsObj *>(currTest[0].get());
-    printf("first train element:\nname: %s\n", example1->imgName.c_str());
+    TR_charsObj *exampleValidation = static_cast<TR_charsObj *>(currValidation[0].get());
-    printf("label: %u\n", example1->label);
+    printf("first train element:\nname: %s\n", exampleTrain->imgName.c_str());
-    printf("first test element:\nname: %s\n", example2->imgName.c_str());
+    printf("label: %u\n", exampleTrain->label);
-    printf("label: %u\n", example2->label);
+    printf("first test element:\nname: %s\n", exampleTest->imgName.c_str());
+    printf("label: %u\n", exampleTest->label);
+    printf("first validation element:\nname: %s\n", exampleValidation->imgName.c_str());
+    printf("label: %u\n", exampleValidation->label);
 
     return 0;
 }

modules/datasetstools/src/gr_chalearn.cpp

@@ -60,6 +60,8 @@ public:
 
 private:
     void loadDataset(const std::string &path);
+
+    void loadDatasetPart(const std::string &path, std::vector< Ptr<Object> > &dataset_, bool loadLabels);
 };
 
 /*GR_chalearnImp::GR_chalearnImp(const string &path)
@@ -77,7 +79,7 @@ void GR_chalearnImp::load(const string &path, int number)
     loadDataset(path);
 }
 
-void GR_chalearnImp::loadDataset(const string &path)
+void GR_chalearnImp::loadDatasetPart(const string &path, vector< Ptr<Object> > &dataset_, bool loadLabels)
 {
     vector<string> fileNames;
     getDirList(path, fileNames);
@@ -101,6 +103,8 @@ void GR_chalearnImp::loadDataset(const string &path)
         curr->depth = atoi(elems[2].c_str());
 
         // loading ground truth
+        if (loadLabels)
+        {
             string fileGroundTruth(path + curr->name + "/" + curr->name + "_labels.csv");
             ifstream infileGroundTruth(fileGroundTruth.c_str());
             while (getline(infileGroundTruth, line))
@@ -115,6 +119,7 @@ void GR_chalearnImp::loadDataset(const string &path)
 
                 curr->groundTruths.push_back(currGroundTruth);
             }
+        }
 
         // loading skeleton
         string fileSkeleton(path + curr->name + "/" + curr->name + "_skeleton.csv");
@@ -142,8 +147,18 @@ void GR_chalearnImp::loadDataset(const string &path)
             curr->skeletons.push_back(currSkeleton);
         }
 
-        train.push_back(curr);
+        dataset_.push_back(curr);
+    }
 }
+
+void GR_chalearnImp::loadDataset(const string &path)
+{
+    string pathTrain(path + "Train/");
+    loadDatasetPart(pathTrain, train, true);
+
+    // freely available validation set doesn't have labels
+    string pathValidation(path + "Validation/");
+    loadDatasetPart(pathValidation, validation, false);
 }
 
 Ptr<GR_chalearn> GR_chalearn::create()

modules/datasetstools/src/gr_skig.cpp

@@ -100,18 +100,28 @@ void GR_skigImp::loadDataset(const string &path)
             curr->dep[0] = 'K';
             curr->dep = pathDatasetDep + curr->dep;
 
-            size_t pos = file.find("person_"); // TODO: check ::npos
+            size_t posPerson = file.find("person_");
-            curr->person = (unsigned char)atoi( file.substr(pos+strlen("person_"), 1).c_str() );
+            size_t posBackground = file.find("backgroud_");
-            pos = file.find("backgroud_");
+            size_t posIllumination = file.find("illumination_");
-            curr->background = (backgroundType)atoi( file.substr(pos+strlen("backgroud_"), 1).c_str() );
+            size_t posPose = file.find("pose_");
-            pos = file.find("illumination_");
+            size_t posType = file.find("actionType_");
-            curr->illumination = (illuminationType)atoi( file.substr(pos+strlen("illumination_"), 1).c_str() );
+            if (string::npos != posPerson &&
-            pos = file.find("pose_");
+                string::npos != posBackground &&
-            curr->pose = (poseType)atoi( file.substr(pos+strlen("pose_"), 1).c_str() );
+                string::npos != posIllumination &&
-            pos = file.find("actionType_");
+                string::npos != posPose &&
-            curr->type = (actionType)atoi( file.substr(pos+strlen("actionType_"), 2).c_str() );
+                string::npos != posType)
+            {
+                curr->person = (unsigned char)atoi( file.substr(posPerson + strlen("person_"), 1).c_str() );
+                curr->background = (backgroundType)atoi( file.substr(posBackground + strlen("backgroud_"), 1).c_str() );
+                curr->illumination = (illuminationType)atoi( file.substr(posIllumination + strlen("illumination_"), 1).c_str() );
+                curr->pose = (poseType)atoi( file.substr(posPose + strlen("pose_"), 1).c_str() );
+                curr->type = (actionType)atoi( file.substr(posType + strlen("actionType_"), 2).c_str() );
 
                 train.push_back(curr);
+            } else
+            {
+                printf("incorrect file name: %s", file.c_str());
+            }
         }
     }
 }

modules/datasetstools/src/ir_robot.cpp

@@ -89,9 +89,20 @@ void IR_robotImp::loadDataset(const string &path)
         string pathScene(path + curr->name + "/");
         vector<string> sceneNames;
         getDirList(pathScene, sceneNames);
+        int currImageNum = 0;
         for (vector<string>::iterator itScene=sceneNames.begin(); itScene!=sceneNames.end(); ++itScene)
         {
-            curr->images.push_back(*itScene);
+            string &fileName = *itScene;
+
+            int imageNum = atoi( fileName.substr(3, 3).c_str() );
+            int pos = atoi( fileName.substr(6, 2).c_str() );
+            if (imageNum != currImageNum)
+            {
+                curr->pos.push_back(cameraPos());
+                currImageNum = imageNum;
+            }
+
+            curr->pos.back().images.push_back(fileName);
         }
 
         train.push_back(curr);

modules/datasetstools/src/msm_epfl.cpp

@@ -60,20 +60,8 @@ public:
 
 private:
     void loadDataset(const std::string &path);
-
-    void readFileDouble(const std::string &fileName, std::vector<double> &out);
 };
 
-void MSM_epflImp::readFileDouble(const string &fileName, vector<double> &out)
-{
-    ifstream infile(fileName.c_str());
-    double val;
-    while (infile >> val)
-    {
-        out.push_back(val);
-    }
-}
-
 /*MSM_epflImp::MSM_epflImp(const string &path)
 {
     loadDataset(path);
@@ -103,9 +91,58 @@ void MSM_epflImp::loadDataset(const string &path)
         Ptr<MSM_epflObj> curr(new MSM_epflObj);
         curr->imageName = *it;
 
-        readFileDouble(string(pathBounding + curr->imageName + ".bounding"), curr->bounding);
+        // load boundary
-        readFileDouble(string(pathCamera + curr->imageName + ".camera"), curr->camera);
+        string fileBounding(pathBounding + curr->imageName + ".bounding");
-        readFileDouble(string(pathP + curr->imageName + ".P"), curr->p);
+        ifstream infile(fileBounding.c_str());
+        for (int k=0; k<2; ++k)
+        {
+            for (int j=0; j<3; ++j)
+            {
+                infile >> curr->bounding(k, j);
+            }
+        }
+
+        // load camera parameters
+        string fileCamera(pathCamera + curr->imageName + ".camera");
+        ifstream infileCamera(fileCamera.c_str());
+        for (int i=0; i<3; ++i)
+        {
+            for (int j=0; j<3; ++j)
+            {
+                infileCamera >> curr->camera.mat1(i, j);
+            }
+        }
+
+        for (int i=0; i<3; ++i)
+        {
+            infileCamera >> curr->camera.mat2[i];
+        }
+
+        for (int i=0; i<3; ++i)
+        {
+            for (int j=0; j<3; ++j)
+            {
+                infileCamera >> curr->camera.mat3(i, j);
+            }
+        }
+
+        for (int i=0; i<3; ++i)
+        {
+            infileCamera >> curr->camera.mat4[i];
+        }
+
+        infileCamera >> curr->camera.imageWidth >> curr->camera.imageHeight;
+
+        // load P
+        string fileP(pathP + curr->imageName + ".P");
+        ifstream infileP(fileP.c_str());
+        for (int k=0; k<3; ++k)
+        {
+            for (int j=0; j<4; ++j)
+            {
+                infileP >> curr->p(k, j);
+            }
+        }
 
         train.push_back(curr);
     }

modules/datasetstools/src/slam_tumindoor.cpp

@@ -81,8 +81,34 @@ void SLAM_tumindoorImp::load(const string &path, int number)
 
 void SLAM_tumindoorImp::loadDataset(const string &path)
 {
-    string infoPath(path + "info/2011-12-17_15.02.56-info.csv"); // TODO
+    string infoPath(path + "info/");
-    ifstream infile(infoPath.c_str());
+
+    // get info map name, .csv should be only one such file in folder
+    string csvName;
+    vector<string> infoNames;
+    getDirList(infoPath, infoNames);
+    for (vector<string>::iterator it=infoNames.begin(); it!=infoNames.end(); ++it)
+    {
+        string &name = *it;
+        if (name.length()>3 && name.substr( name.length()-4, 4 )==".csv")
+        {
+            if (csvName.length()==0)
+            {
+                csvName = name;
+            } else
+            {
+                printf("more than one .csv file in info folder\n");
+                return;
+            }
+        }
+    }
+    if (csvName.length()==0)
+    {
+        printf("didn't find .csv file in info folder\n");
+        return;
+    }
+
+    ifstream infile((infoPath + csvName).c_str());
     string line;
     while (getline(infile, line))
     {

modules/datasetstools/src/tr_chars.cpp

@@ -53,15 +53,19 @@ class CV_EXPORTS TR_charsImp : public TR_chars
 {
 public:
     TR_charsImp() {}
-    //TR_charsImp(const std::string &path, int number = 0);
+    //TR_charsImp(const string &path, int number = 0);
     virtual ~TR_charsImp() {}
 
-    virtual void load(const std::string &path, int number = 0);
+    virtual void load(const string &path, int number = 0);
 
 private:
-    void loadDataset(const std::string &path, int number = 0);
+    void loadDataset(const string &path, int number = 0);
 
-    void parseLine(const std::string &line, std::vector<int> &currSet, int number);
+    void parseLine(const string &line, vector<int> &currSet, int number);
+
+    inline void convert(vector<int> &from, vector< Ptr<Object> > &to, vector<int> &allLabels, vector<string> &allNames);
+
+    inline void parseSet(const string &line, const string &pattern, bool &flag, vector<int> &set, int number);
 };
 
 void TR_charsImp::parseLine(const string &line, vector<int> &currSet, int number)
@@ -80,6 +84,38 @@ void TR_charsImp::parseLine(const string &line, vector<int> &currSet, int number
     }
 }
 
+inline void TR_charsImp::convert(vector<int> &from, std::vector< Ptr<Object> > &to, vector<int> &allLabels, vector<string> &allNames)
+{
+    for (vector<int>::iterator it=from.begin(); it!=from.end(); ++it)
+    {
+        if (*it>=(int)allNames.size() || *it>=(int)allLabels.size())
+        {
+            printf("incorrect index: %u\n", *it);
+            continue;
+        }
+
+        Ptr<TR_charsObj> curr(new TR_charsObj);
+        curr->imgName = allNames[*it];
+        curr->label = allLabels[*it];
+        to.push_back(curr);
+    }
+}
+
+inline void TR_charsImp::parseSet(const string &line, const string &pattern, bool &flag, vector<int> &set, int number)
+{
+    size_t pos = line.find(pattern);
+    if (string::npos != pos)
+    {
+        flag = true;
+        string s(line.substr(pos + pattern.length()));
+        parseLine(s, set, number);
+    } else
+    if (flag)
+    {
+        parseLine(line, set, number);
+    }
+}
+
 /*TR_charsImp::TR_charsImp(const string &path, int number)
 {
     loadDataset(path, number);
@@ -92,12 +128,12 @@ void TR_charsImp::load(const string &path, int number)
 
 void TR_charsImp::loadDataset(const string &path, int number)
 {
-    vector<int> allLabels, trainSet, testSet;
+    vector<int> allLabels, trainSet, testSet, validationSet;
     vector<string> allNames;
 
     ifstream infile((path + "list_English_Img.m").c_str());
     string line;
-    bool labels = false, names = false, isTrain = false, isTest = false;
+    bool labels = false, names = false, isTrain = false, isTest = false, isValidation = false;
     while (getline(infile, line))
     {
         size_t pos = line.find("];");
@@ -107,6 +143,7 @@ void TR_charsImp::loadDataset(const string &path, int number)
             names = false;
             isTrain = false;
             isTest = false;
+            isValidation = false;
         }
 
         string slabels("list.ALLlabels = [");
@@ -137,66 +174,24 @@ void TR_charsImp::loadDataset(const string &path, int number)
             allNames.push_back(s);
         }
 
-        string strain("list.TRNind = [");
+        string trainStr("list.TRNind = [");
-        pos = line.find(strain);
+        parseSet(line, trainStr, isTrain, trainSet, number);
-        if (string::npos!=pos)
-        {
-            isTrain = true;
-            string s(line.substr(pos+strain.length()));
-            parseLine(s, trainSet, number);
-        } else
-        if (isTrain)
-        {
-            parseLine(line, trainSet, number);
-        }
 
-        string stest("list.TSTind = [");
+        string testStr("list.TSTind = [");
-        pos = line.find(stest);
+        parseSet(line, testStr, isTest, testSet, number);
-        if (string::npos!=pos)
+
-        {
+        string validationStr("list.VALind = [");
-            isTest = true;
+        parseSet(line, validationStr, isValidation, validationSet, number);
-            string s(line.substr(pos+stest.length()));
-            parseLine(s, testSet, number);
-        } else
-        if (isTest)
-        {
-            parseLine(line, testSet, number);
-        }
 
         /*"list.classlabels = ["
         "list.classnames = ["
         "list.NUMclasses = 62;"
-        "list.VALind = [" // TODO: load validation
         "list.TXNind = ["*/
     }
 
-    for (vector<int>::iterator it=trainSet.begin(); it!=trainSet.end(); ++it)
+    convert(trainSet, train, allLabels, allNames);
-    {
+    convert(testSet, test, allLabels, allNames);
-        if (*it>=(int)allNames.size() || *it>=(int)allLabels.size())
+    convert(validationSet, validation, allLabels, allNames);
-        {
-            printf("incorrect train index: %u\n", *it);
-            continue;
-        }
-
-        Ptr<TR_charsObj> curr(new TR_charsObj);
-        curr->imgName = allNames[*it];
-        curr->label = allLabels[*it];
-        train.push_back(curr);
-    }
-
-    for (vector<int>::iterator it=testSet.begin(); it!=testSet.end(); ++it)
-    {
-        if (*it>=(int)allNames.size() || *it>=(int)allLabels.size())
-        {
-            printf("incorrect test index: %u\n", *it);
-            continue;
-        }
-
-        Ptr<TR_charsObj> curr(new TR_charsObj);
-        curr->imgName = allNames[*it];
-        curr->label = allLabels[*it];
-        test.push_back(curr);
-    }
 }
 
 Ptr<TR_chars> TR_chars::create()