#include <QCoreApplication>
#include "bufferfile.h"
#include <opencv2/opencv.hpp>
#include <QDebug>
#include <cblas.h>
#include <time.h>
void getFeature(PredictorHandle &pred_hnd,
std::vector<mx_float> &image_data,
mx_uint image_size,
float *feature)
{
// // Set Input Image
MXPredSetInput(pred_hnd, "data", image_data.data(), image_size);
// // Do Predict Forward
MXPredForward(pred_hnd);
mx_uint output_index = 0;
mx_uint *shape = 0;
mx_uint shape_len;
mx_uint size = 1;
// Get Output Result
MXPredGetOutputShape(pred_hnd, output_index, &shape, &shape_len);
for (mx_uint i = 0; i < shape_len; ++i) { size *= shape[i]; }
int64 start = cv::getTickCount();
MXPredGetOutput(pred_hnd, output_index, feature, size);
int64 end = cv::getTickCount();
double secs = (end-start)/cv::getTickFrequency();
// qDebug()<<"time: " <<secs;
}
void GetImageFile(const std::string image_file,
mx_float* image_data,
const uint channels,
const cv::Size resize_size,
const mx_float* mean_data = nullptr) {
// Read all kinds of file into a BGR color 3 channels image
cv::Mat im_ori = cv::imread(image_file, cv::IMREAD_COLOR);
if (im_ori.empty()) {
std::cerr << "Can't open the image. Please check " << image_file << ". \n";
assert(false);
}
cv::Mat im;
cv::resize(im_ori, im, resize_size);//缩放图像
// cv::imshow(image_file,im);
// cv::waitKey(1);
// float mean[3]={104.0,117.0,123.0};
float mean[3]={0};
// float mean[3]={128.0};
int b = 0,g = 1, r = 2;
int w = im.cols;
int h = im.rows;
for(int i=0; i < h; ++i){
uchar* data = im.ptr<uchar>(i);
for(int j=0; j< w; ++j){
image_data[2*w*h+i*w+j] = static_cast<mx_float>(*data++);//- mean[b];
image_data[w*h+i*w+j] = static_cast<mx_float>(*data++);//- mean[g];
image_data[i*w+j] = static_cast<mx_float>(*data++);//- mean[r];
}
}
}
int test()
{
// model file path
std::string json_file = "E:/GitHub/Python2/InsightFace/insightface/models/model-r100-ii/model-symbol.json";
std::string param_file = "E:/GitHub/Python2/InsightFace/insightface/models/model-r100-ii/model-0000.params";
// read model file
BufferFile json_data(json_file);
BufferFile param_data(param_file);
if (json_data.GetLength() == 0 || param_data.GetLength() == 0)
{
return EXIT_FAILURE;
}
// mxnet parameters
int dev_type = 2; // 1: cpu, 2: gpu, we can change
int dev_id = 0; // arbitrary.
mx_uint num_input_nodes = 1; // 1 for feedforward
const char *input_key[1] = { "data" };
const char **input_keys = input_key;
uint width = 112;
uint height = 112;
uint channels = 3;
const mx_uint input_shape_indptr[2] = { 0, 4}; // column dim is 2
const mx_uint input_shape_data[4] = { 1,channels,height,width};
// global predicator handler
PredictorHandle pred_hnd = nullptr;
// create predictor
MXPredCreate(static_cast<const char*>(json_data.GetBuffer()),
static_cast<const char*>(param_data.GetBuffer()),
static_cast<int>(param_data.GetLength()),
dev_type,
dev_id,
num_input_nodes,
input_keys,
input_shape_indptr,
input_shape_data,
&pred_hnd);
if (!pred_hnd)
{
std::cerr << "Failed to create predict handler" << std::endl;
return EXIT_FAILURE;
}
mx_uint image_size = width * height * channels;
std::string test_file1 = "D:/IMG/video/0.jpg";
// // Read Image Data
std::vector<mx_float> image_data1 = std::vector<mx_float>(image_size);
GetImageFile(test_file1, image_data1.data(), channels, cv::Size(112, 112));
float f1[512] ={0};
getFeature(pred_hnd,image_data1,image_size,&f1[0]);
std::string test_file2 = "D:/IMG/video/";
for(int i =0;i<=7;i++)
{
std::string path;
path.append(test_file2).append(std::to_string(i)).append(".jpg");
float f2[512] ={0};
std::vector<mx_float> image_data2 = std::vector<mx_float>(image_size);
GetImageFile(path, image_data2.data(), channels, cv::Size(112, 112));
getFeature(pred_hnd,image_data2,image_size,&f2[0]);
float sim = cblas_sdot(512,&f1[0],1,&f2[0],1)/
(sqrt(cblas_sdot(512,&f1[0],1,&f1[0],1))*sqrt(cblas_sdot(512,&f2[0],1,&f2[0],1)));
// sim = 0.5+0.5*sim;//归一化
qDebug()<<test_file1.c_str()<<"vs"<<test_file2.c_str()<<sim;
}
return EXIT_SUCCESS;
}
int main(int argc, char *argv[])
{
QCoreApplication a(argc, argv);
test();
return a.exec();
}