added benchmark for ar_hmdb

modules/datasets/CMakeLists.txt

@@ -1,3 +1,4 @@
 set(the_description "datasets framework")
-ocv_define_module(datasets opencv_core opencv_face)
+ocv_define_module(datasets opencv_core opencv_face opencv_ml opencv_flann)
 
+ocv_warnings_disable(CMAKE_CXX_FLAGS /wd4267) # flann, Win64

modules/datasets/doc/datasets/ar_hmdb.rst

@@ -14,3 +14,16 @@ _`"HMDB: A Large Human Motion Database"`: http://serre-lab.clps.brown.edu/resour
 
  3. To load data run: ./opencv/build/bin/example_datasets_ar_hmdb -p=/home/user/path_to_unpacked_folders/
 
+Benchmark
+"""""""""
+
+For this dataset was implemented benchmark, which gives accuracy: 0.107407 (using precomputed HOG/HOF "STIP" features from site, averaging for 3 splits)
+
+To run this benchmark execute:
+
+.. code-block:: bash
+
+ ./opencv/build/bin/example_datasets_ar_hmdb_benchmark -p=/home/user/path_to_unpacked_folders/
+
+(precomputed features should be unpacked in the same folder: /home/user/path_to_unpacked_folders/hmdb51_org_stips/)
+

modules/datasets/doc/datasets/fr_lfw.rst

@@ -14,8 +14,14 @@ _`"Labeled Faces in the Wild"`: http://vis-www.cs.umass.edu/lfw/
 
  3. To load data run: ./opencv/build/bin/example_datasets_fr_lfw -p=/home/user/path_to_unpacked_folder/lfw2/
 
-.. note:: Benchmark
+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_datasets_fr_lfw_benchmark -p=/home/user/path_to_unpacked_folder/lfw2/
+For this dataset was implemented benchmark, which gives accuracy: 0.623833 +- 0.005223 (train split: pairsDevTrain.txt, dataset: lfwa)
+
+To run this benchmark execute:
+
+.. code-block:: bash
+
+ ./opencv/build/bin/example_datasets_fr_lfw_benchmark -p=/home/user/path_to_unpacked_folder/lfw2/
 

modules/datasets/samples/ar_hmdb_benchmark.cpp

@@ -0,0 +1,276 @@
+/*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/datasets/ar_hmdb.hpp"
+#include "opencv2/datasets/util.hpp"
+
+#include <opencv2/core.hpp>
+#include <opencv2/flann.hpp>
+#include <opencv2/ml.hpp>
+
+#include <cstdio>
+
+#include <string>
+#include <vector>
+#include <fstream>
+
+using namespace std;
+using namespace cv;
+using namespace cv::datasets;
+using namespace cv::flann;
+using namespace cv::ml;
+
+unsigned int getNumFiles(vector< Ptr<Object> > &curr);
+unsigned int getNumFiles(vector< Ptr<Object> > &curr)
+{
+    unsigned int numFiles = 0;
+    for (unsigned int i=0; i<curr.size(); ++i)
+    {
+        AR_hmdbObj *example = static_cast<AR_hmdbObj *>(curr[i].get());
+        vector<string> &videoNames = example->videoNames;
+        for (vector<string>::iterator it=videoNames.begin(); it!=videoNames.end(); ++it)
+        {
+            numFiles++;
+        }
+    }
+
+    return numFiles;
+}
+
+void fillData(const string &path, vector< Ptr<Object> > &curr, Index &flann_index, Mat1f &data, Mat1i &labels);
+void fillData(const string &path, vector< Ptr<Object> > &curr, Index &flann_index, Mat1f &data, Mat1i &labels)
+{
+    const unsigned int descriptorNum = 162;
+    Mat1f sample(1, descriptorNum);
+    Mat1i nresps(1, 1);
+    Mat1f dists(1, 1);
+
+    unsigned int numFiles = 0;
+    for (unsigned int i=0; i<curr.size(); ++i)
+    {
+        AR_hmdbObj *example = static_cast<AR_hmdbObj *>(curr[i].get());
+        vector<string> &videoNames = example->videoNames;
+        for (vector<string>::iterator it=videoNames.begin(); it!=videoNames.end(); ++it)
+        {
+            string featuresFile = *it + ".txt";
+            string featuresFullPath = path + "hmdb51_org_stips/" + example->name + "/" + featuresFile;
+
+            ifstream infile(featuresFullPath.c_str());
+            string line;
+            // skip header
+            for (unsigned int j=0; j<3; ++j)
+            {
+                getline(infile, line);
+            }
+            while (getline(infile, line))
+            {
+                // 7 skip, hog+hof: 72+90 read
+                vector<string> elems;
+                split(line, elems, '\t');
+
+                for (unsigned int j=0; j<descriptorNum; ++j)
+                {
+                    sample(0, j) = (float)atof(elems[j+7].c_str());
+                }
+
+                flann_index.knnSearch(sample, nresps, dists, 1, SearchParams());
+                data(numFiles, nresps(0, 0)) ++;
+            }
+            labels(numFiles, 0) = i;
+            numFiles++;
+        }
+    }
+}
+
+int main(int argc, char *argv[])
+{
+    const char *keys =
+            "{ help h usage ? |    | show this message }"
+            "{ path p         |true| path to dataset }";
+    CommandLineParser parser(argc, argv, keys);
+    string path(parser.get<string>("path"));
+    if (parser.has("help") || path=="true")
+    {
+        parser.printMessage();
+        return -1;
+    }
+
+    // loading dataset
+    Ptr<AR_hmdb> dataset = AR_hmdb::create();
+    dataset->load(path);
+
+    int numSplits = dataset->getNumSplits();
+    printf("splits number: %u\n", numSplits);
+
+
+    const unsigned int descriptorNum = 162;
+    const unsigned int clusterNum = 4000;
+    const unsigned int sampleNum = 5613856; // max for all 3 splits
+
+    vector<double> res;
+    for (int currSplit=0; currSplit<numSplits; ++currSplit)
+    {
+
+        Mat1f samples(sampleNum, descriptorNum);
+        unsigned int currSample = 0;
+        vector< Ptr<Object> > &curr = dataset->getTrain(currSplit);
+        unsigned int numTrainFiles = getNumFiles(curr);
+        unsigned int numFeatures = 0;
+        for (unsigned int i=0; i<curr.size(); ++i)
+        {
+            AR_hmdbObj *example = static_cast<AR_hmdbObj *>(curr[i].get());
+            vector<string> &videoNames = example->videoNames;
+            for (vector<string>::iterator it=videoNames.begin(); it!=videoNames.end(); ++it)
+            {
+                string featuresFile = *it + ".txt";
+                string featuresFullPath = path + "hmdb51_org_stips/" + example->name + "/" + featuresFile;
+
+                ifstream infile(featuresFullPath.c_str());
+                string line;
+                // skip header
+                for (unsigned int j=0; j<3; ++j)
+                {
+                    getline(infile, line);
+                }
+                while (getline(infile, line))
+                {
+                    numFeatures++;
+                    if (currSample < sampleNum)
+                    {
+                        // 7 skip, hog+hof: 72+90 read
+                        vector<string> elems;
+                        split(line, elems, '\t');
+
+                        for (unsigned int j=0; j<descriptorNum; ++j)
+                        {
+                            samples(currSample, j) = (float)atof(elems[j+7].c_str());
+                        }
+                        currSample++;
+                    }
+                }
+            }
+        }
+        printf("split %u, train features number: %u, samples number: %u\n", currSplit, numFeatures, currSample);
+
+        // clustering
+        Mat1f centers(clusterNum, descriptorNum);
+        ::cvflann::KMeansIndexParams kmean_params;
+        unsigned int resultClusters = hierarchicalClustering< L2<float> >(samples, centers, kmean_params);
+        if (resultClusters < clusterNum)
+        {
+            centers = centers.rowRange(Range(0, resultClusters));
+        }
+        Index flann_index(centers, KDTreeIndexParams());
+        printf("resulted clusters number: %u\n", resultClusters);
+
+
+        Mat1f trainData(numTrainFiles, resultClusters);
+        Mat1i trainLabels(numTrainFiles, 1);
+
+        for (unsigned int i=0; i<numTrainFiles; ++i)
+        {
+            for (unsigned int j=0; j<resultClusters; ++j)
+            {
+                trainData(i, j) = 0;
+            }
+        }
+
+        printf("calculating train histograms\n");
+        fillData(path, curr, flann_index, trainData, trainLabels);
+
+        printf("train svm\n");
+        SVM::Params params;
+        params.svmType = SVM::C_SVC;
+        params.kernelType = SVM::POLY; //SVM::RBF;
+        params.degree = 0.5;
+        params.gamma = 1;
+        params.coef0 = 1;
+        params.C = 1;
+        params.nu = 0.5;
+        params.p = 0;
+        params.termCrit = TermCriteria(TermCriteria::MAX_ITER+TermCriteria::EPS, 1000, 0.01);
+        Ptr<SVM> svm = SVM::create(params);
+        svm->train(trainData, ROW_SAMPLE, trainLabels);
+
+        // prepare to predict
+        curr = dataset->getTest(currSplit);
+        unsigned int numTestFiles = getNumFiles(curr);
+        Mat1f testData(numTestFiles, resultClusters);
+        Mat1i testLabels(numTestFiles, 1); // ground true
+
+        for (unsigned int i=0; i<numTestFiles; ++i)
+        {
+            for (unsigned int j=0; j<resultClusters; ++j)
+            {
+                testData(i, j) = 0;
+            }
+        }
+
+        printf("calculating test histograms\n");
+        fillData(path, curr, flann_index, testData, testLabels);
+
+        printf("predicting\n");
+        Mat1f testPredicted(numTestFiles, 1);
+        svm->predict(testData, testPredicted);
+
+        unsigned int correct = 0;
+        for (unsigned int i=0; i<numTestFiles; ++i)
+        {
+            if ((int)testPredicted(i, 0) == testLabels(i, 0))
+            {
+                correct++;
+            }
+        }
+        double accuracy = 1.0*correct/numTestFiles;
+        printf("correctly recognized actions: %f\n", accuracy);
+        res.push_back(accuracy);
+
+    }
+
+    double accuracy = 0.0;
+    for (unsigned int i=0; i<res.size(); ++i)
+    {
+        accuracy += res[i];
+    }
+    printf("average: %f\n", accuracy/res.size());
+
+    return 0;
+}