Qiea/c-cnn
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

实现cnn计算功能

Browse Source
This commit is contained in:
Qiea
2024-11-11 11:42:08 +08:00
parent 034e3a5302
commit fe062deba5
1 changed files with 15 additions and 37 deletions
Download Patch File Download Diff File Expand all files Collapse all files

52
cnn.c
View File

@@ -62,10 +62,10 @@ float* convolution(Model model_w, Model model_b, const float* input_matrix, int
for(int k=0; k < model_w.num_kernels; k++) {
for (int row = 0; row < cr_l; row++) {
for (int col = 0; col < cr_l; col++) {
conv_temp = 0; // 每个输出像素初始化为0
for (int c = 0; c < model_w.channel; c++) {
conv_temp += _conv_rlst[(c * model_w.num_kernels * cr_l * cr_l) + (k * cr_l * cr_l) + (row * cr_l + col)];
}
conv_rlst[(k * cr_l * cr_l) + (row * cr_l + col)] = conv_temp;
// 加上对应卷积核的偏置
conv_temp += model_b.array[k];
// 激活函数：ReLU（将小于0的值设为0）
@@ -77,27 +77,6 @@ float* convolution(Model model_w, Model model_b, const float* input_matrix, int
}
}
}
// if(model_w.channel>1){
// float sum=0;
// for(int _l=0 ; _l < model_w.channel ; _l++){
// sum += input_matrix[
// (_l*(im_l*im_l)) + ((row+1)*(im_l)+(col+1))
// ];
// }
// conv_temp = sum;
// }
//
// if(_debug < 10 && strcmp(model_w.name, "conv2_weight") == 0){
// printf("[%d]%f, input_matrix:[%f]\r\k",_debug,conv_temp,input_matrix[
// (0*(im_l*im_l)) + ((row+1)*(im_l)+(col+1))
// ]);
// }
// _debug++;
return conv_rlst;
}
@@ -174,9 +153,9 @@ int main(){
//第二层：填充32 * 52 * 52
float* expand_matrix_2 = expand(pool_rslt_1, 50, 32);
print_rslt(expand_matrix_2, 52, (1*10*52));
// print_rslt(expand_matrix_2, 52, (1*10*52));
float* conv_rlst_2 = convolution(conv2_weight,conv2_bias,expand_matrix_2, 52);
print_rslt(conv_rlst_2, 50, (1*10*50));
// print_rslt(conv_rlst_2, 50, (64*50*50));
float* pool_rslt_2 = pooling(conv2_weight, conv_rlst_2, 50);
// print_rslt(pool_rslt_2, 25, (1*25*25));
@@ -184,7 +163,7 @@ int main(){
float* expand_matrix_3 = expand(pool_rslt_2, 25, 64);
// print_rslt(expand_matrix_2, 52, (1*52*52));
float* conv_rlst_3 = convolution(conv3_weight,conv3_bias,expand_matrix_3, 27);
// print_rslt(conv_rlst_3, 25, (1*25*25));
print_rslt(conv_rlst_3, 25, (1*25*25));
float* pool_rslt_3 = pooling(conv3_weight, conv_rlst_3, 25);
// print_rslt(pool_rslt_3, 12, (1*12*12));
@@ -228,31 +207,29 @@ int main(){
affine1_rslt[n] = 0; // 否则存入0
}
// print_rslt(affine1_rslt,1,128);
print_rslt(affine1_rslt,1,128);
float affine2_temp; // 临时变量，用于存储输出层的中间结果
float* affine2_rslt = (float *) malloc(sizeof(float)*7);
memset(affine2_rslt, 0, 7);
float affine2_rslt[7]; // 存储输出层的结果（假设输出层有7个神经元）
// 比较输出层的最大值
float temp = -100; // 用于存储最大值的临时变量，初始化为一个非常小的值
u8 predict_num; // 用于存储预测的数字（对应最大值的索引）
int predict_num; // 用于存储预测的数字（对应最大值的索引）
// 遍历7个输出神经元（假设有7个类别）
for(u8 n=0; n<7; n++)
// 遍历10个输出神经元（假设有10个类别）
for(int n=0; n<7; n++)
{
affine2_temp = 0; // 当前神经元的输出初始化为0
// 进行矩阵乘法，将隐藏层的输出与输出层权重进行点积
for(u8 i=0; i<128; i++)
for(int i=0; i<128; i++)
{
affine2_temp = affine2_temp + fc2_weight.array[i+100*n] * affine1_rslt[i];
affine2_temp = affine2_temp + fc2_weight.array[i+128*n] * affine1_rslt[i];
}
// 加上对应神经元的偏置
affine2_temp = affine2_temp + fc2_bias.array[n];
affine2_temp = affine2_temp + fc2_weight.array[n];
affine2_rslt[n] = affine2_temp; // 存储输出层的结果
// 寻找最大值
@@ -263,8 +240,9 @@ int main(){
}
}
print_rslt(affine2_rslt,7,7);
printf("RES is:%d",predict_num+1);
print_rslt(affine2_rslt,1,7);
printf("Label is:%d",predict_num+1);
return 0;
}