• Home
  • History
  • Annotate
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*//////////////////////////////////////////////////////////////////////////////////////
2 // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
3 
4 //  By downloading, copying, installing or using the software you agree to this license.
5 //  If you do not agree to this license, do not download, install,
6 //  copy or use the software.
7 
8 // This is a implementation of the Logistic Regression algorithm in C++ in OpenCV.
9 
10 // AUTHOR:
11 // Rahul Kavi rahulkavi[at]live[at]com
12 //
13 
14 // contains a subset of data from the popular Iris Dataset (taken from
15 // "http://archive.ics.uci.edu/ml/datasets/Iris")
16 
17 // # You are free to use, change, or redistribute the code in any way you wish for
18 // # non-commercial purposes, but please maintain the name of the original author.
19 // # This code comes with no warranty of any kind.
20 
21 // #
22 // # You are free to use, change, or redistribute the code in any way you wish for
23 // # non-commercial purposes, but please maintain the name of the original author.
24 // # This code comes with no warranty of any kind.
25 
26 // # Logistic Regression ALGORITHM
27 
28 //                           License Agreement
29 //                For Open Source Computer Vision Library
30 
31 // Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
32 // Copyright (C) 2008-2011, Willow Garage Inc., all rights reserved.
33 // Third party copyrights are property of their respective owners.
34 
35 // Redistribution and use in source and binary forms, with or without modification,
36 // are permitted provided that the following conditions are met:
37 
38 //   * Redistributions of source code must retain the above copyright notice,
39 //     this list of conditions and the following disclaimer.
40 
41 //   * Redistributions in binary form must reproduce the above copyright notice,
42 //     this list of conditions and the following disclaimer in the documentation
43 //     and/or other materials provided with the distribution.
44 
45 //   * The name of the copyright holders may not be used to endorse or promote products
46 //     derived from this software without specific prior written permission.
47 
48 // This software is provided by the copyright holders and contributors "as is" and
49 // any express or implied warranties, including, but not limited to, the implied
50 // warranties of merchantability and fitness for a particular purpose are disclaimed.
51 // In no event shall the Intel Corporation or contributors be liable for any direct,
52 // indirect, incidental, special, exemplary, or consequential damages
53 // (including, but not limited to, procurement of substitute goods or services;
54 // loss of use, data, or profits; or business interruption) however caused
55 // and on any theory of liability, whether in contract, strict liability,
56 // or tort (including negligence or otherwise) arising in any way out of
57 // the use of this software, even if advised of the possibility of such damage.*/
58 
59 #include <iostream>
60 
61 #include <opencv2/core.hpp>
62 #include <opencv2/ml.hpp>
63 #include <opencv2/highgui.hpp>
64 
65 using namespace std;
66 using namespace cv;
67 using namespace cv::ml;
68 
showImage(const Mat & data,int columns,const String & name)69 static void showImage(const Mat &data, int columns, const String &name)
70 {
71     Mat bigImage;
72     for(int i = 0; i < data.rows; ++i)
73     {
74         bigImage.push_back(data.row(i).reshape(0, columns));
75     }
76     imshow(name, bigImage.t());
77 }
78 
calculateAccuracyPercent(const Mat & original,const Mat & predicted)79 static float calculateAccuracyPercent(const Mat &original, const Mat &predicted)
80 {
81     return 100 * (float)countNonZero(original == predicted) / predicted.rows;
82 }
83 
main()84 int main()
85 {
86     const String filename = "../data/data01.xml";
87     cout << "**********************************************************************" << endl;
88     cout << filename
89          << " contains digits 0 and 1 of 20 samples each, collected on an Android device" << endl;
90     cout << "Each of the collected images are of size 28 x 28 re-arranged to 1 x 784 matrix"
91          << endl;
92     cout << "**********************************************************************" << endl;
93 
94     Mat data, labels;
95     {
96         cout << "loading the dataset...";
97         FileStorage f;
98         if(f.open(filename, FileStorage::READ))
99         {
100             f["datamat"] >> data;
101             f["labelsmat"] >> labels;
102             f.release();
103         }
104         else
105         {
106             cerr << "file can not be opened: " << filename << endl;
107             return 1;
108         }
109         data.convertTo(data, CV_32F);
110         labels.convertTo(labels, CV_32F);
111         cout << "read " << data.rows << " rows of data" << endl;
112     }
113 
114     Mat data_train, data_test;
115     Mat labels_train, labels_test;
116     for(int i = 0; i < data.rows; i++)
117     {
118         if(i % 2 == 0)
119         {
120             data_train.push_back(data.row(i));
121             labels_train.push_back(labels.row(i));
122         }
123         else
124         {
125             data_test.push_back(data.row(i));
126             labels_test.push_back(labels.row(i));
127         }
128     }
129     cout << "training/testing samples count: " << data_train.rows << "/" << data_test.rows << endl;
130 
131     // display sample image
132     showImage(data_train, 28, "train data");
133     showImage(data_test, 28, "test data");
134 
135     // simple case with batch gradient
136     cout << "training...";
137     //! [init]
138     Ptr<LogisticRegression> lr1 = LogisticRegression::create();
139     lr1->setLearningRate(0.001);
140     lr1->setIterations(10);
141     lr1->setRegularization(LogisticRegression::REG_L2);
142     lr1->setTrainMethod(LogisticRegression::BATCH);
143     lr1->setMiniBatchSize(1);
144     //! [init]
145     lr1->train(data_train, ROW_SAMPLE, labels_train);
146     cout << "done!" << endl;
147 
148     cout << "predicting...";
149     Mat responses;
150     lr1->predict(data_test, responses);
151     cout << "done!" << endl;
152 
153     // show prediction report
154     cout << "original vs predicted:" << endl;
155     labels_test.convertTo(labels_test, CV_32S);
156     cout << labels_test.t() << endl;
157     cout << responses.t() << endl;
158     cout << "accuracy: " << calculateAccuracyPercent(labels_test, responses) << "%" << endl;
159 
160     // save the classfier
161     const String saveFilename = "NewLR_Trained.xml";
162     cout << "saving the classifier to " << saveFilename << endl;
163     lr1->save(saveFilename);
164 
165     // load the classifier onto new object
166     cout << "loading a new classifier from " << saveFilename << endl;
167     Ptr<LogisticRegression> lr2 = StatModel::load<LogisticRegression>(saveFilename);
168 
169     // predict using loaded classifier
170     cout << "predicting the dataset using the loaded classfier...";
171     Mat responses2;
172     lr2->predict(data_test, responses2);
173     cout << "done!" << endl;
174 
175     // calculate accuracy
176     cout << labels_test.t() << endl;
177     cout << responses2.t() << endl;
178     cout << "accuracy: " << calculateAccuracyPercent(labels_test, responses2) << "%" << endl;
179 
180     waitKey(0);
181     return 0;
182 }
183