added benchmark for fr_lfw dataset

modules/datasetstools/CMakeLists.txt

 set(the_description "datasets tools")
-ocv_define_module(datasetstools opencv_core)
+ocv_define_module(datasetstools opencv_core opencv_face)
 

modules/datasetstools/doc/datasetstools.rst

@@ -52,12 +52,17 @@ _`"Labeled Faces in the Wild"`: http://vis-www.cs.umass.edu/lfw/
 
 .. note:: Usage
 
- 1. From link above download any dataset file: lfw.tgz\lfwa.tar.gz\lfw-deepfunneled.tgz\lfw-funneled.tgz and file with 10 test splits: pairs.txt.
+ 1. From link above download any dataset file: lfw.tgz\lfwa.tar.gz\lfw-deepfunneled.tgz\lfw-funneled.tgz and files with pairs: with 10 test splits: pairs.txt and with developer train split: pairsDevTrain.txt.
 
- 2. Unpack dataset file and place pairs.txt in created folder.
+ 2. Unpack dataset file and place pairs.txt and pairsDevTrain.txt in created folder.
 
  3. To load data run: ./opencv/build/bin/example_datasetstools_fr_lfw -p=/home/user/path_to_unpacked_folder/lfw2/
 
+.. note:: Benchmark
+
+ - For this dataset was implemented benchmark, which gives accuracy: 0.623833 +- 0.005223 (train split: pairsDevTrain.txt, dataset: lfwa)
+ - To run this benchmark execute: ./opencv/build/bin/example_datasetstools_fr_lfw_benchmark -p=/home/user/path_to_unpacked_folder/lfw2/
+
 Gesture Recognition
 -------------------
 

modules/datasetstools/include/opencv2/datasetstools/util.hpp

@@ -52,9 +52,9 @@ namespace cv
 namespace datasetstools
 {
 
-void split(const std::string &s, std::vector<std::string> &elems, char delim);
+void CV_EXPORTS split(const std::string &s, std::vector<std::string> &elems, char delim);
 
-void getDirList(const std::string &dirName, std::vector<std::string> &fileNames);
+void CV_EXPORTS getDirList(const std::string &dirName, std::vector<std::string> &fileNames);
 
 }
 }

modules/datasetstools/samples/fr_lfw_benchmark.cpp

+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2014, Itseez Inc, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Itseez Inc or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#include "opencv2/core.hpp"
+#include "opencv2/imgcodecs.hpp"
+
+#include "opencv2/face.hpp"
+#include "opencv2/datasetstools/fr_lfw.hpp"
+
+#include <iostream>
+
+#include <cstdio>
+
+#include <string>
+#include <vector>
+#include <map>
+
+using namespace std;
+using namespace cv;
+using namespace cv::datasetstools;
+using namespace cv::face;
+
+map<string, int> people;
+
+int getLabel(const string &imagePath);
+int getLabel(const string &imagePath)
+{
+    size_t pos = imagePath.find('/');
+    string curr = imagePath.substr(0, pos);
+    map<string, int>::iterator it = people.find(curr);
+    if (people.end() == it)
+    {
+        people.insert(make_pair(curr, people.size()));
+        it = people.find(curr);
+    }
+    return (*it).second;
+}
+
+int main(int argc, const char *argv[])
+{
+    const char *keys =
+            "{ help h usage ? |    | show this message }"
+            "{ path p         |true| path to dataset (lfw2 folder) }";
+    CommandLineParser parser(argc, argv, keys);
+    string path(parser.get<string>("path"));
+    if (parser.has("help") || path=="true")
+    {
+        parser.printMessage();
+        return -1;
+    }
+
+    // These vectors hold the images and corresponding labels.
+    vector<Mat> images;
+    vector<int> labels;
+
+    // load dataset
+    Ptr<FR_lfw> dataset = FR_lfw::create();
+    dataset->load(path);
+
+    unsigned int numSplits = dataset->getNumSplits();
+    printf("splits number: %u\n", numSplits);
+    printf("train size: %u\n", (unsigned int)dataset->getTrain().size());
+    printf("test size: %u\n", (unsigned int)dataset->getTest().size());
+
+    for (unsigned int i=0; i<dataset->getTrain().size(); ++i)
+    {
+        FR_lfwObj *example = static_cast<FR_lfwObj *>(dataset->getTrain()[i].get());
+
+        int currNum = getLabel(example->image1);
+        Mat img = imread(path+example->image1, IMREAD_GRAYSCALE);
+        images.push_back(img);
+        labels.push_back(currNum);
+
+        currNum = getLabel(example->image2);
+        img = imread(path+example->image2, IMREAD_GRAYSCALE);
+        images.push_back(img);
+        labels.push_back(currNum);
+    }
+
+    // 2200 pairsDevTrain, first split: correct: 373, from: 600 -> 0.621667%
+    Ptr<FaceRecognizer> model = createLBPHFaceRecognizer();
+    // 2200 pairsDevTrain, first split: correct: correct: 369, from: 600 -> 0.615%
+    //Ptr<FaceRecognizer> model = createEigenFaceRecognizer();
+    // 2200 pairsDevTrain, first split: correct: 372, from: 600 -> 0.62%
+    //Ptr<FaceRecognizer> model = createFisherFaceRecognizer();
+
+    model->train(images, labels);
+    //string saveModelPath = "face-rec-model.txt";
+    //cout << "Saving the trained model to " << saveModelPath << endl;
+    //model->save(saveModelPath);
+
+    vector<double> p;
+    for (unsigned int j=0; j<numSplits; ++j)
+    {
+        unsigned int incorrect = 0, correct = 0;
+        vector < Ptr<Object> > &curr = dataset->getTest(j);
+        for (unsigned int i=0; i<curr.size(); ++i)
+        {
+            FR_lfwObj *example = static_cast<FR_lfwObj *>(curr[i].get());
+
+            //int currNum = getLabel(example->image1);
+            Mat img = imread(path+example->image1, IMREAD_GRAYSCALE);
+            int predictedLabel1 = model->predict(img);
+
+            //currNum = getLabel(example->image2);
+            img = imread(path+example->image2, IMREAD_GRAYSCALE);
+            int predictedLabel2 = model->predict(img);
+
+            if ((predictedLabel1 == predictedLabel2 && example->same) ||
+                (predictedLabel1 != predictedLabel2 && !example->same))
+            {
+                correct++;
+            } else
+            {
+                incorrect++;
+            }
+        }
+        p.push_back(1.0*correct/(correct+incorrect));
+        printf("correct: %u, from: %u -> %f%%\n", correct, correct+incorrect, p.back());
+    }
+    double mu = 0.0;
+    for (vector<double>::iterator it=p.begin(); it!=p.end(); ++it)
+    {
+        mu += *it;
+    }
+    mu /= p.size();
+    double sigma = 0.0;
+    for (vector<double>::iterator it=p.begin(); it!=p.end(); ++it)
+    {
+        sigma += (*it - mu)*(*it - mu);
+    }
+    sigma = sqrt(sigma/p.size());
+    double se = sigma/sqrt(p.size());
+    printf("estimated mean accuracy: %f and the standard error of the mean: %f\n", mu, se);
+
+    return 0;
+}

modules/datasetstools/samples/or_mnist.cpp

@@ -57,7 +57,7 @@ int main(int argc, char *argv[])
 {
     const char *keys =
             "{ help h usage ? |    | show this message }"
-            "{ path p         |true| path to dataset (SUN397 folder) }";
+            "{ path p         |true| path to dataset }";
     CommandLineParser parser(argc, argv, keys);
     string path(parser.get<string>("path"));
     if (parser.has("help") || path=="true")

modules/datasetstools/src/fr_lfw.cpp

@@ -82,12 +82,13 @@ void FR_lfwImp::loadDataset(const string &path)
     getDirList(path, fileNames);
     for (vector<string>::iterator it=fileNames.begin(); it!=fileNames.end(); ++it)
     {
-        if ("pairs.txt" == *it)
+        string &name = *it;
+
+        if (name.length()>3 && name.substr(name.length()-4) == ".txt")
         {
             continue;
         }
 
-        string &name = *it;
         vector<string> images;
 
         string pathFace(path + name + "/");
@@ -142,6 +143,37 @@ void FR_lfwImp::loadDataset(const string &path)
 
         num++;
     }
+
+    // dev train loading to train[0]
+    ifstream infile2((path + "pairsDevTrain.txt").c_str());
+    getline(infile2, line); // should 1100
+    while (getline(infile2, line))
+    {
+        vector<string> elems;
+        split(line, elems, '\t');
+
+        Ptr<FR_lfwObj> curr(new FR_lfwObj);
+        string &person1 = elems[0];
+        unsigned int imageNumber1 = atoi(elems[1].c_str())-1;
+        curr->image1 = person1 + "/" + faces[person1][imageNumber1];
+
+        string person2;
+        unsigned int imageNumber2;
+        if (3 == elems.size())
+        {
+            person2 = elems[0];
+            imageNumber2 = atoi(elems[2].c_str())-1;
+            curr->same = true;
+        } else
+        {
+            person2 = elems[2];
+            imageNumber2 = atoi(elems[3].c_str())-1;
+            curr->same = false;
+        }
+        curr->image2 = person2 + "/" + faces[person2][imageNumber2];
+
+        train[0].push_back(curr);
+    }
 }
 
 Ptr<FR_lfw> FR_lfw::create()