-
Notifications
You must be signed in to change notification settings - Fork 4
/
randomtest.h
263 lines (214 loc) · 7.77 KB
/
randomtest.h
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
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
#ifndef RANDOMTEST_H_
#define RANDOMTEST_H_
#include "data.h"
#include "utilities.h"
#include <cmath>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
using namespace cv;
using namespace std;
class RandomTest {
public:
RandomTest() {
}
RandomTest(const int &numClasses) :
m_numClasses(&numClasses), m_trueCount(0.0), m_falseCount(0.0),xyzR(3,vector<float>(0)),xyzL(3,vector<float>(0)),thitaR(3,vector<float>(0)),thitaL(3,vector<float>(0)) {
for (int i = 0; i <numClasses; i++) {
m_trueStats.push_back(0.0);
m_falseStats.push_back(0.0);
}
}
void updateStats(const Sample &sample, const bool decision) {
if (decision) {
m_trueCount += sample.w;
m_trueStats[sample.ma] += sample.w;
for(int i=0;i<3;i++){
xyzR[i].push_back(sample.y[i]);
thitaR[i].push_back(sample.y[i+3]);
}
} else {
m_falseCount += sample.w;
m_falseStats[sample.ma] += sample.w;
for(int i=0;i<3;i++){
xyzL[i].push_back(sample.y[i]);
thitaL[i].push_back(sample.y[i+3]);
}
}
}
double score(DataSet &dataset_tr) {
double totalCount = m_trueCount + m_falseCount;
double p=1/totalCount;
double p1,p2,splitEntropy = 0.0;
if (m_trueCount) {
p1 = m_trueCount / totalCount;
splitEntropy -= p1 * log(p)/log(2.0);
}
if (m_falseCount) {
p2 = m_falseCount / totalCount;
splitEntropy -= p2 * log(p)/log(2.0);
}
double p3=1/m_trueCount, p4=1/m_falseCount;
vector<double>m_totalStats(*m_numClasses,0.0);
for (int i = 0; i < *m_numClasses; i++)
m_totalStats[i] = (m_trueStats[i] + m_falseStats[i]);
double total=0;
for(int i = 0; i < *m_numClasses; i++)
total+= m_totalStats[i];
double Entropytotal=0;
for(int i = 0; i < *m_numClasses; i++)
if(m_totalStats[i])
{
Entropytotal-= (m_totalStats[i]/total)*log((m_totalStats[i]/total))/log(2.0);
}
int counterR=0,counterL=0;
int rowR=0,rowL=0;
float avgxyzR[3]={0,0,0},avgxyzL[3]={0,0,0},avgthitaR[3]={0,0,0},avgthitaL[3]={0,0,0};
for(int i=0;i<m_trueCount;i++)
{
avgxyzR[0]+=xyzR[0][i];
avgxyzR[1]+=xyzR[1][i];
avgxyzR[2]+=xyzR[2][i];
avgthitaR[0]+=thitaR[0][i];
avgthitaR[1]+=thitaR[0][i];
avgthitaR[2]+=thitaR[0][i];
}
for(int i=0;i<m_falseCount;i++)
{
avgxyzL[0]+=xyzL[0][i];
avgxyzL[1]+=xyzL[1][i];
avgxyzL[2]+=xyzL[2][i];
avgthitaL[0]+=thitaL[0][i];
avgthitaL[1]+=thitaL[0][i];
avgthitaL[2]+=thitaL[0][i];
}
for(int i=0;i<3;i++)
{
if(m_trueCount)
{
avgxyzR[i]=avgxyzR[i]/m_trueCount;
avgthitaR[i]=avgthitaR[i]/m_trueCount;
}
if(m_falseCount)
{
avgxyzL[i]=avgxyzL[i]/m_falseCount;
avgthitaL[i]=avgthitaL[i]/m_falseCount;
}
}
float covxyzR[3][3]={0},covthitaR[3][3]={0},covxyzL[3][3]={0},covthitaL[3][3]={0};
for(int i=0,j=0;i<3,j<3;i++,j++)
for(int q=0;q<m_trueCount;q++)
{
covxyzR[i][j]+=(xyzR[i][q]-avgxyzR[i])*(xyzR[j][q]-avgxyzR[j]);
covthitaR[i][j]+=(thitaR[i][q]-avgthitaR[i])*(thitaR[j][q]-avgthitaR[j]);
}
if(m_trueCount)
for(int i=0;i<3;i++)
for(int j=0;j<3;j++)
{
covxyzR[i][j]=covxyzR[i][j]/m_trueCount;
covthitaR[i][j]=covthitaR[i][j]/m_trueCount;
}
for(int i=0,j=0;i<3,j<3;i++,j++)
for(int q=0;q<m_falseCount;q++)
{
covxyzL[i][j]+=(xyzL[i][q]-avgxyzL[i])*(xyzL[j][q]-avgxyzL[j]);
covthitaL[i][j]+=(thitaL[i][q]-avgthitaL[i])*(thitaL[j][q]-avgthitaL[j]);
}
if(m_falseCount)
for(int i=0;i<3;i++)
for(int j=0;j<3;j++)
{
covxyzL[i][j]=covxyzL[i][j]/m_falseCount;
covthitaL[i][j]=covthitaL[i][j]/m_falseCount;
}
float hlsxyzR=sqrt(covxyzR[0][0]*covxyzR[1][1]*covxyzR[2][2])+sqrt(covxyzR[0][1]*covxyzR[1][2]*covxyzR[2][0])+sqrt(covxyzR[0][2]*covxyzR[1][0]*covxyzR[2][1])-sqrt(covxyzR[0][2]*covxyzR[1][1]*covxyzR[2][0])-sqrt(covxyzR[0][0]*covxyzR[1][2]*covxyzR[2][1])-sqrt(covxyzR[0][1]*covxyzR[1][0]*covxyzR[2][2]);
float hlsxyzL=sqrt(covxyzL[0][0]*covxyzL[1][1]*covxyzL[2][2])+sqrt(covxyzL[0][1]*covxyzL[1][2]*covxyzL[2][0])+sqrt(covxyzL[0][2]*covxyzL[1][0]*covxyzL[2][1])-sqrt(covxyzL[0][2]*covxyzL[1][1]*covxyzL[2][0])-sqrt(covxyzL[0][0]*covxyzL[1][2]*covxyzL[2][1])-sqrt(covxyzL[0][1]*covxyzL[1][0]*covxyzL[2][2]);
float hlsthitaR=sqrt(covthitaR[0][0]*covthitaR[1][1]*covthitaR[2][2])+sqrt(covthitaR[0][1]*covthitaR[1][2]*covthitaR[2][0])+sqrt(covthitaR[0][2]*covthitaR[1][0]*covthitaR[2][1])-sqrt(covthitaR[0][2]*covthitaR[1][1]*covthitaR[2][0])-sqrt(covthitaR[0][0]*covthitaR[1][2]*covthitaR[2][1])-sqrt(covthitaR[0][1]*covthitaR[1][0]*covthitaR[2][2]);
float hlsthitaL=sqrt(covthitaL[0][0]*covthitaL[1][1]*covthitaL[2][2])+sqrt(covthitaL[0][1]*covthitaL[1][2]*covthitaL[2][0])+sqrt(covthitaL[0][2]*covthitaL[1][0]*covthitaL[2][1])-sqrt(covthitaL[0][2]*covthitaL[1][1]*covthitaL[2][0])-sqrt(covthitaL[0][0]*covthitaL[1][2]*covthitaL[2][1])-sqrt(covthitaL[0][1]*covthitaL[1][0]*covthitaL[2][2]);
double result=Entropytotal-p3*(hlsxyzR+hlsthitaR)-p4*(hlsxyzL+hlsthitaL);
return result;
}
pair<vector<double> , vector<double> > getStats() {
return pair<vector<double> , vector<double> > (m_trueStats, m_falseStats);
}
protected:
const int *m_numClasses;
double m_threshold;
double m_trueCount;
double m_falseCount;
vector<double> m_trueStats;
vector<double> m_falseStats;
vector<vector<float>> xyzR;
vector<vector<float>> xyzL;
vector<vector<float>> thitaR;
vector<vector<float>> thitaL;
};
class HyperplaneFeature: public RandomTest {
public:
HyperplaneFeature() {
}
HyperplaneFeature(const int &numClasses) :
RandomTest(numClasses),test_point(20) {
}
int test_point;
vector<double> dif_squ_rec;
void update(Sample &sample) {
updateStats(sample, eval(sample));
}
bool eval(Sample &sample) {
dif_squ_rec.resize(test_point/2);
int p_width=80,p_height=80;
int side=5;
int q=0;
vector<vector<double>> patches(p_width,vector<double>(p_height));
for (int j=0;j<p_width;j++)
for (int k=0;k<p_height;k++){
patches[j][k]=sample.x[q];
q++;
}
vector<Rect> points(test_point);
for (int j=0;j<test_point;j++)
{
points[j].x=(int)randomFromRange(0,p_width-side);
points[j].y=(int)randomFromRange(0,p_height-side);
points[j].width=side;
points[j].height=side;
}
vector<vector<vector<int>>> test_rec(test_point,vector<vector<int>>(side,vector<int>(side)));
for (int j=0;j<test_point;j++)
for(int l=0;l<side;l++)
for(int m=0;m<side;m++)
test_rec[j][l][m]=patches[points[j].x+l][points[j].y+m];//i表示第i个patch，j若为双数，则表示第（j+2）/2对矩形中的第一个矩形
vector<vector<vector<int>>> dif_rec(test_point/2,vector<vector<int>>(side,vector<int>(side)));
for (int j=0;j<test_point;j=j+2)
for(int l=0;l<side;l++)
for(int m=0;m<side;m++)
dif_rec[j/2][l][m]=test_rec[j][l][m]-test_rec[j+1][l][m];
for (int j=0;j<test_point/2;j++)
dif_squ_rec[j]=0;
for (int j=0;j<test_point/2;j++)
for(int l=0;l<side;l++)
for(int m=0;m<side;m++)
dif_squ_rec[j]+=sqrt((double)dif_rec[j][l][m]*dif_rec[j][l][m]);
int indexj=0;
int maxvalue=0;
int minvalue=0;
for (int j=0;j<test_point/2;j++)
{
if(dif_squ_rec[j]>maxvalue)
{
maxvalue=dif_squ_rec[j];
indexj=j;
}
if(dif_squ_rec[j]<minvalue)
{
minvalue=dif_squ_rec[j];
}
}
m_threshold = randomFromRange(minvalue, maxvalue);
int testsel=(int)randomFromRange(0,test_point/2);
double proj=dif_squ_rec[testsel];
return (proj >=m_threshold) ? true : false;
}
};
#endif /* RANDOMTEST_H_ */