-
Notifications
You must be signed in to change notification settings - Fork 0
/
CNN_CPP.cpp
241 lines (214 loc) · 9.61 KB
/
CNN_CPP.cpp
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
//
// Created by 任艺伟 on 2020/12/19.
//
#include "CNN_CPP.h"
picture::picture(Mat image) {
Mat BGR[3];
resize(image, image, Size(128, 128));
split(image,BGR);
this->picture_size = image.rows;
this->channels = image.channels();
this->pixel = new float[picture_size * picture_size * channels]();
int k = 0;
for(int i=0;i<BGR[2].total();i++){
pixel[k++] = (float)BGR[2].data[i]/255.0f;
}//R
for(int i=0;i<BGR[1].total();i++){
pixel[k++] = (float)BGR[1].data[i]/255.0f;
}//G
for(int i=0;i<BGR[0].total();i++){
pixel[k++] = (float)BGR[0].data[i]/255.0f;
}//B
counter = new atomic_int;
*(counter) = 1;
}
picture::picture() {
counter = new atomic_int;
*(counter) = 0;
}
picture::picture(int picture_size, int channels, float *pixel) {
this->channels = channels;
this->picture_size = picture_size;
this->pixel = new float[picture_size * picture_size * channels]();
for (int i = 0; i < picture_size * picture_size * channels; i++) {
*(this->pixel + i) = pixel[i];
}
this->counter = new atomic_int;
*(counter) = 1;
}
Matrix picture::pictureToMatrix(int kernel_size, int channel, int stride, int padding) {
// cout<<this->channels<<" "<<this->pixel[0]<<endl;
int out_size = floor((this->picture_size - kernel_size + 2 * padding) / stride + 1);
Matrix matrix(out_size * out_size, kernel_size * kernel_size * channel+1);
int current_position = 0;
// bool judge = (this->picture_size - kernel_size + 2 * padding) % stride == 0;
// int step = judge ? picture_size - kernel_size + padding : picture_size - kernel_size + padding-1;//有问题
// for (int m = 0; m <= step; m += m > 0 ? stride : m - padding + stride) {
// for (int n = 0; n <= step; n += n > 0 ? stride : n - padding + stride) {
// for (int i = 0; i < this->channels; i++) {
// if (padding > m) current_position += (padding - m) * kernel_size;
// for (int j = 0; j < kernel_size; j++) {
// if (padding <= m &&m<=picture_size - kernel_size|| padding > m
// && j < kernel_size - padding||m>picture_size - kernel_size&&m+j<picture_size) {
// if (padding > n) current_position +=padding - n;
// for (int k = 0; k < kernel_size; k++) {
// if (padding <= n&&n<=picture_size - kernel_size|| padding > n
// && k < kernel_size - padding||n>picture_size - kernel_size&&n+k<picture_size)
// matrix.getValue()[current_position++] = this->pixel[n + m*picture_size +i * picture_size *picture_size +j * picture_size + k];
// else if(n+k>=kernel_size) current_position+= (n+k-picture_size+1)*1;
// }
// }else if(m+j>=picture_size) current_position+= (m+j-picture_size+1)*kernel_size;
// }
// }
// }
// }
//
// for(int m=0;m+kernel_size<=picture_size+2*padding;m+=stride){
// for(int n=0;n+kernel_size<=picture_size+2*padding;n+=stride){
// for(int i=0;i<channel;i++){
// for(int j=0;j<kernel_size;j++){
// for(int k=0;k<kernel_size;k++){
// if(m+j-padding>=0&&n+k-padding>=0&&m+j<padding+picture_size&&n+k<picture_size+padding){
// matrix.getValue()[current_position++]
// = this->pixel[i*picture_size*picture_size+picture_size*(m+j-padding)+n+k-padding];
// }else current_position++;
// }
// }
// }
// }
// }
for(int m=0;m+kernel_size<=picture_size+2*padding;m+=stride){
for(int n=0;n+kernel_size<=picture_size+2*padding;n+=stride){
for(int i=0;i<channel;i++){
for(int j=0;j<kernel_size;j++){
for(int k=0;k<kernel_size;k++){
if(m+j-padding>=0&&n+k-padding>=0&&m+j<padding+picture_size&&n+k<picture_size+padding){
matrix.getValue()[current_position++] = this->pixel[i*picture_size*picture_size+picture_size*(m+j-padding)+n+k-padding];
}else current_position++;
}
}
}
matrix.getValue()[current_position++] = 1;
}
}
return matrix;
}
Matrix convToMatrix(const conv_param& currentKernel) {
Matrix conv_result(currentKernel.kernel_size * currentKernel.kernel_size * currentKernel.in_channels+1,currentKernel.out_channels);
int currentPosition = 0;
for (int k = 0; k <currentKernel.in_channels* currentKernel.kernel_size * currentKernel.kernel_size; k++) {
for (int i = 0; i < currentKernel.out_channels; i++) {
conv_result.getValue()[currentPosition++] = currentKernel.p_weight[
i * currentKernel.kernel_size * currentKernel.kernel_size * currentKernel.in_channels + k];
}
}
for (int i = 0; i < currentKernel.out_channels; i++) {
conv_result.getValue()[currentPosition++] = currentKernel.p_bias[i];
}
return conv_result;
}
Matrix BiasToMatrix(const float * bias,int out_size,int out_channels){
Matrix result(out_channels,out_size*out_size);
int currentPosition = 0;
for(int j =0;j<out_channels;j++){
for(int i=0;i<out_size*out_size;i++){
result.getValue()[currentPosition++] = bias[j];
}
}
return result;
}
void picture::ConBVReLU(const conv_param ¤tKernel) {
Matrix p_matrix = pictureToMatrix(currentKernel.kernel_size,currentKernel.in_channels,currentKernel.stride,currentKernel.pad);
// p_matrix.transposition();
int out_size = floor((this->picture_size - currentKernel.kernel_size + 2 * currentKernel.pad) / currentKernel.stride + 1);
Matrix c_matrix = convToMatrix(currentKernel);
// c_matrix.transposition();
// Matrix CBR_matrix(currentKernel.out_channels,out_size*out_size);
// Matrix b_matrix = BiasToMatrix(currentKernel.p_bias,out_size,currentKernel.out_channels);
// cblas_sgemm(CblasRowMajor, CblasTrans, CblasTrans, c_matrix.getColumnNumber(),
// p_matrix.getRowNumber(), c_matrix.getRowNumber(), 1.0, c_matrix.getValue(), c_matrix.getColumnNumber(),
// p_matrix.getValue(), p_matrix.getColumnNumber(), 0.0, CBR_matrix.getValue(), p_matrix.getRowNumber());
// cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, c_matrix.getRowNumber(), p_matrix.getColumnNumber(), c_matrix.getColumnNumber(), 1.0, c_matrix.getValue(), c_matrix.getColumnNumber(), p_matrix.getValue(), p_matrix.getColumnNumber(), 0.0, CBR_matrix.getValue(), p_matrix.getColumnNumber());
// cout<<p_matrix.getRowNumber()<<" "<<p_matrix.getColumnNumber()<<endl;
// cout<<c_matrix.getRowNumber()<<" "<<c_matrix.getColumnNumber()<<endl;
// cout<<b_matrix.getRowNumber()<<" "<<b_matrix.getColumnNumber()<<endl;
// Matrix CBR_matrix = (p_matrix*c_matrix)+b_matrix;
Matrix CBR_matrix = p_matrix*c_matrix;
this->picture_size = out_size;
this->channels = currentKernel.out_channels;
if(*(counter) == 1) {
delete counter;
delete [] this->pixel ;
}else if(*(counter)==0){
delete counter;
}
else (*counter)--;
this->pixel = CBR_matrix.getValue();
this->counter = CBR_matrix.getCounter();
(*this->counter)++;
}
void picture::MaxPooling(int stride, int wide) {
int out_size = this->picture_size%stride==0?picture_size/stride:this->picture_size / stride + 1;
float * afterPolling= new float[out_size*out_size*channels]();
int currentPosition = 0;
for (int k = 0; k < channels; k++) {
for (int m = 0; m < picture_size; m += stride) {
for (int n = 0; n < picture_size; n += stride) {
float resultValue = 0;
for (int i = 0; i < wide; i++) {
if (m + i < picture_size)
for (int j = 0; j < wide; j++) {
if (n + j < picture_size) resultValue =
max(resultValue, pixel[k*picture_size*picture_size+(m+i)*picture_size+n+j]);
}
}
afterPolling[currentPosition++] = resultValue;
}
}
}
if(*(counter) == 1) {
delete counter;
delete [] this->pixel ;
}else if(*(counter)==0){
delete counter;
}
else (*counter)--;
this->picture_size = out_size;
this->pixel = afterPolling;
this->counter = new atomic_int;
*(counter) = 1;
}
void picture::FullyConnected(const fc_param &fc) {
Matrix weight(fc.out_features,fc.in_features,fc.p_weight,1);
weight.setCounter(2);
Matrix pictureNow(picture_size*picture_size*channels,1,pixel,1);
pictureNow.setCounter(counter);
(*counter)++;
Matrix res = weight*pictureNow;
for(int i=0;i<fc.out_features;i++){
res.getValue()[i] += fc.p_bias[i];
}
float sum=0;
for(int i=0;i<fc.out_features;i++){
sum += exp(res.getValue()[i]);
// cout << res.getValue()[i] << endl;
}
for(int i = 0;i<fc.out_features;i++){
res.getValue()[i] = exp(res.getValue()[i])/sum;
cout<<res.getValue()[i]<<endl;
if(i==1){
if(res.getValue()[1]>0.97) cout<<"你很像人！！"<<endl;
else if(res.getValue()[1]>0.7) cout<<"你可能是个人"<<endl;
else if(res.getValue()[1]<0.1) cout<<"你必不是人"<<endl;
}
}
}
picture::~picture() {
if(*(counter) == 1) {
delete counter;
delete [] this->pixel ;
}else if(*(counter)==0){
delete counter;
}
else (*counter)--;
}