[OpenCV] Samples 06: logistic regression

logistic regression，这个算法只能解决简单的线性二分类，在众多的机器学习分类算法中并不出众，但它能被改进为多分类，并换了另外一个名字softmax, 这可是深度学习中响当当的分类算法。 

 

Reference: denny的学习专栏  // 臭味相投的一个博客

 

  • Xml保存图片的方法和读取的方式。
  • Mat显示内部的多个图片。
  • Mat::t() 显示矩阵内容。

 

本文用它来进行手写数字分类。

在opencv3.0中提供了一个xml文件，里面存放了40个样本，分别是20个数字0的手写体和20个数字1的手写体。本来每个数字的手写体是一张28*28的小图片，在xml使用1*784 的向量保存在<data>中。

这个文件的位置： \opencv\sources\samples\data\data01.xml

 

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
/*//////////////////////////////////////////////////////////////////////////////////////
// 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.
 
// This is a implementation of the Logistic Regression algorithm in C++ in OpenCV.
 
// AUTHOR:
// Rahul Kavi rahulkavi[at]live[at]com
//
 
// contains a subset of data from the popular Iris Dataset (taken from
// "http://archive.ics.uci.edu/ml/datasets/Iris")
 
// # You are free to use, change, or redistribute the code in any way you wish for
// # non-commercial purposes, but please maintain the name of the original author.
// # This code comes with no warranty of any kind.
 
// #
// # You are free to use, change, or redistribute the code in any way you wish for
// # non-commercial purposes, but please maintain the name of the original author.
// # This code comes with no warranty of any kind.
 
// # Logistic Regression ALGORITHM
 
//                           License Agreement
//                For Open Source Computer Vision Library
 
// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
// Copyright (C) 2008-2011, Willow Garage 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:
 
//   * Redistributions of source code must retain the above copyright notice,
//     this list of conditions and the following disclaimer.
 
//   * Redistributions 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 Intel Corporation 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.*/
 
#include <iostream>
 
#include <opencv2/core.hpp>
#include <opencv2/ml.hpp>
#include <opencv2/highgui.hpp>
 
using namespace std;
using namespace cv;
using namespace cv::ml;
 
 
/*
 * Jeff --> Show mutiple-photos from Mat.
 */
static void showImage(const Mat &data, int columns, const String &name)
{
    // columns = 28
    Mat bigImage;
    for(int i = 0; i < data.rows; ++i)
    {
        //rows: number of photos.
        // vector --> reshape --> col 28, col 28 ...
        // push_back: show each pic from left to right.
        bigImage.push_back(data.row(i).reshape(0, columns));
 
    }
    imshow(name, bigImage.t());
}
 
static float calculateAccuracyPercent(const Mat &original, const Mat &predicted)
{
    return 100 * (float)countNonZero(original == predicted) / predicted.rows;
}
 
int main()
{
    const String filename = "../data/data01.xml";
    cout << "**********************************************************************" << endl;
    cout << filename
         << " contains digits 0 and 1 of 20 samples each, collected on an Android device" << endl;
    cout << "Each of the collected images are of size 28 x 28 re-arranged to 1 x 784 matrix"
         << endl;
    cout << "**********************************************************************" << endl;
 
    Mat data, labels;
    {
        /*
         * Jeff --> Load xml.
         *          transform to Mat.
         * FileStorage.
         */
        cout << "loading the dataset...";
        // Step 1.
        FileStorage f;
        if(f.open(filename, FileStorage::READ))
        {
            // Step 2.
            f["datamat"] >> data;
            f["labelsmat"] >> labels;
            f.release();
        }
        else
        {
            cerr << "file can not be opened: " << filename << endl;
            return 1;
        }
        // Step 3.
        data.convertTo(data, CV_32F);
        labels.convertTo(labels, CV_32F);
 
        cout << "read " << data.rows << " rows of data" << endl;
    }
 
    Mat data_train, data_test;
    Mat labels_train, labels_test;
    for(int i = 0; i < data.rows; i++)
    {
        // Step 4.
        if(i % 2 == 0)
        {
            data_train.push_back(data.row(i));
            labels_train.push_back(labels.row(i));
        }
        else
        {
            data_test.push_back(data.row(i));
            labels_test.push_back(labels.row(i));
        }
    }
    cout << "training/testing samples count: " << data_train.rows << "/" << data_test.rows << endl;
 
    // display sample image
    showImage(data_train, 28, "train data");
    showImage(data_test, 28, "test data");
 
    /**************************************************************************/
 
    // simple case with batch gradient
    cout << "training...";
 
    // Step (1), create classifier.
    Ptr<LogisticRegression> lr1 = LogisticRegression::create();
 
    // Step (2),
    lr1->setLearningRate(0.001);
    lr1->setIterations(10);
    lr1->setRegularization(LogisticRegression::REG_L2);
    lr1->setTrainMethod(LogisticRegression::BATCH);
    lr1->setMiniBatchSize(1);
 
    // Step (3), train.
    //! [init]
    lr1->train(data_train, ROW_SAMPLE, labels_train);
    cout << "done!" << endl;
 
    //--------------------------------------------------------------------------
 
    cout << "predicting...";
 
    // Step (4), predict.
    Mat responses;
    lr1->predict(data_test, responses);
    cout << "done!" << endl;
 
 
    // Step (5), show prediction report
    cout << "original vs predicted:" << endl;
    // Jeff --> CV_32S is a signed 32bit integer value for each pixel.
    labels_test.convertTo(labels_test, CV_32S);
 
    cout << labels_test.t() << endl;
    cout << responses.t() << endl;
    cout << "accuracy: " << calculateAccuracyPercent(labels_test, responses) << "%" << endl;
 
    // Step (6), save the classfier
    const String saveFilename = "NewLR_Trained.xml";
    cout << "saving the classifier to " << saveFilename << endl;
    lr1->save(saveFilename);
 
    /****************************** End ***************************************/
 
    // load the classifier onto new object
    cout << "loading a new classifier from " << saveFilename << endl;
    Ptr<LogisticRegression> lr2 = StatModel::load<LogisticRegression>(saveFilename);
 
    // predict using loaded classifier
    cout << "predicting the dataset using the loaded classfier...";
    Mat responses2;
    lr2->predict(data_test, responses2);
    cout << "done!" << endl;
 
    // calculate accuracy
    cout << labels_test.t() << endl;
    cout << responses2.t() << endl;
    cout << "accuracy: " << calculateAccuracyPercent(labels_test, responses2) << "%" << endl;
 
    waitKey(0);
    return 0;
}

 

 

关于逻辑回归：http://blog.csdn.net/pakko/article/details/37878837

 

什么是逻辑回归？

Logistic回归与多重线性回归实际上有很多相同之处，最大的区别就在于它们的因变量不同，其他的基本都差不多。正是因为如此，这两种回归可以归于同一个家族，即广义线性模型（generalizedlinear model）。

这一家族中的模型形式基本上都差不多，不同的就是因变量不同。

  • 如果是连续的，就是多重线性回归；
  • 如果是二项分布，就是Logistic回归；
  • 如果是Poisson分布，就是Poisson回归；
  • 如果是负二项分布，就是负二项回归。

 

Logistic回归的因变量可以是二分类的，也可以是多分类的，但是二分类的更为常用，也更加容易解释。所以实际中最常用的就是二分类的Logistic回归。

Logistic回归的主要用途：

  • 寻找危险因素：寻找某一疾病的危险因素等；
  • 预测：根据模型，预测在不同的自变量情况下，发生某病或某种情况的概率有多大；
  • 判别：实际上跟预测有些类似，也是根据模型，判断某人属于某病或属于某种情况的概率有多大，也就是看一下这个人有多大的可能性是属于某病。

 

Logistic回归主要在流行病学中应用较多，比较常用的情形是探索某疾病的危险因素，根据危险因素预测某疾病发生的概率，等等。例如，想探讨胃癌发生的危险因素，可以选择两组人群，一组是胃癌组，一组是非胃癌组，两组人群肯定有不同的体征和生活方式等。这里的因变量就是是否胃癌，即“是”或“否”，自变量就可以包括很多了，例如年龄、性别、饮食习惯、幽门螺杆菌感染等。自变量既可以是连续的，也可以是分类的。

 

常规步骤

Regression问题的常规步骤为： 

  1. 寻找h函数（即hypothesis）； ==> Sigmoid函数
  2. 构造J函数（loss函数）；
  3. 想办法使得J函数最小并求得回归参数（θ）

 

详见reference博客。

posted @   郝壹贰叁  阅读(1728)  评论(0)  编辑  收藏  举报
点击右上角即可分享
微信分享提示