Linux 硬件加速 开源人脸识别引擎SeetaFace：第二步在ZedBoard上运行
两张图片的相似度主函数
在Linux下运行相似度函数，先输入test_face_verification.cpp的code再进行编译输出结果
#include(gallery_faces.size());
std::vector(probe_faces.size());
if (gallery_face_num == 0 || probe_face_num==0)
{
std::cout << "Faces are not detected.";
return 0;
}
/*3-14for (int32_t i = 0; i < 1; i++) {
//face_rect.x = faces[i].bbox.x;
//face_rect.y = faces[i].bbox.y;
//face_rect.width = faces[i].bbox.width;
//face_rect.height = faces[i].bbox.height;
cout << "x =" << gallery_faces[i].bbox.x << endl;
cout << "y =" << gallery_faces[i].bbox.y << endl;
cout << "w =" << gallery_faces[i].bbox.width << endl;
cout << "h =" << gallery_faces[i].bbox.height << endl;
cout << "x =" << probe_faces[i].bbox.x << endl;
cout << "y =" << probe_faces[i].bbox.y << endl;
cout << "w =" << probe_faces[i].bbox.width << endl;
cout << "h =" << probe_faces[i].bbox.height << endl;
//cv::rectangle(img, face_rect, CV_RGB(0, 0, 255), 4, 8, 0);
}*/
// Detect 5 facial landmarks
seeta::FacialLandmark gallery_points[5];
point_detector.PointDetectLandmarks(gallery_img_data_gray, gallery_faces[0], gallery_points);
seeta::FacialLandmark probe_points[5];
point_detector.PointDetectLandmarks(probe_img_data_gray, probe_faces[0], probe_points);
/* for (int i = 0; i<5; i++)
{
cv::circle(gallery_img_color, cv::Point(gallery_points[i].x, gallery_points[i].y), 2,
CV_RGB(0, 255, 0));
cv::circle(probe_img_color, cv::Point(probe_points[i].x, probe_points[i].y), 2,
CV_RGB(0, 255, 0));
}*/
//cv::imwrite("gallery_point_result.jpg", gallery_img_color);
//cv::imwrite("probe_point_result.jpg", probe_img_color);
// Extract face identity feature
float gallery_fea[2048];
float probe_fea[2048];
/*3-14for(int i = 0; i<5; i++){
std::cout<
face_recognizer.ExtractFeatureWithCrop(gallery_img_data_color, gallery_points, gallery_fea);
face_recognizer.ExtractFeatureWithCrop(probe_img_data_color, probe_points, probe_fea);
// Caculate similarity of two faces
float sim = face_recognizer.CalcSimilarity(gallery_fea, probe_fea);
std::cout << sim << std::endl;
if(sim>mx)
{
mx=sim;
std::cout << mx << std::endl;
temp=i;
std::cout << temp << std::endl;
}
}
//std::cout << mx << std::endl;
if(mx>0.72)
{cout<<" the most likely pic is "<2 *temp]<else
cout<<" The image library does not have the person."<return 0;
} cpan@admin0-PowerEdge-R730:~/workspace/SeetaFaceEngine-master/FaceIdentification/build2$ make -j8
cpan@admin0-PowerEdge-R730:~/workspace/SeetaFaceEngine-master/FaceIdentification/build2$ ./src/test/test_face_verification.bin
经过在Linux环境下测试运行通过将code复制到Xilinx公司的SDSOC SDK运行release生成SD_Card目录下的文件，copy到SD卡上运行，SDSOC上的代码code为：
#include(gallery_faces.size());
std::vector(probe_faces.size());
if (gallery_face_num == 0 || probe_face_num==0)
{
std::cout << "Faces are not detected.";
return 0;
}
// Detect 5 facial landmarks
seeta::FacialLandmark gallery_points[5];
point_detector.PointDetectLandmarks(gallery_img_data_gray, gallery_faces[0], gallery_points);
seeta::FacialLandmark probe_points[5];
point_detector.PointDetectLandmarks(probe_img_data_gray, probe_faces[0], probe_points);
// Extract face identity feature
float gallery_fea[2048];
float probe_fea[2048];
face_recognizer.ExtractFeatureWithCrop(gallery_img_data_color, gallery_points, gallery_fea);
face_recognizer.ExtractFeatureWithCrop(probe_img_data_color, probe_points, probe_fea);
// Caculate similarity of two faces
float sim = face_recognizer.CalcSimilarity(gallery_fea, probe_fea);
std::cout << sim <if(sim>mx)
{
mx=sim;
std::cout << mx << std::endl;
temp=i;
std::cout << temp << std::endl;
}
main_ctr.stop();
std::cout << main_ctr.avg_cpu_cycles() <if(mx>0.7)
{cout<<" the most likely pic is "<2*temp]<else
cout<<" The image library does not have the person."<return 0;
}
程序在经过FPGA加速后运行时间减少一半
Tips：
1、SDSOC： FPGA加速时波特率设置100Mhz matrix_product_hz1
一般刚跑程序时，只需要勾选：Generate SD Card Image 与Estimate Performance
Generate SD Card Image这个将生成的网标文件elf和BIN文件拷贝到板子上运行
Estimate Performance：模拟综合，看资源消耗情况
需要加入FPGA加速的时候，勾选Generate Bitstream，配置FPGA部分，消耗时间长
2、将SDSOC下lab运行处的SD_card所有程序copy到mnt下
运行时将图像库pic放在主目录下，把test与test_g也放在主目录下
运行程序./mnt/SeetaFace.elf
3、由于不用opencv库，所以通过/home/cpan/workspace/SeetaFaceEnginemaster/FaceIdentification/build2下，
运行make -j8
生成test_face_verification.bin文件，在运行该文件
./src/test/test_face_verification.bin
这个是在Linux下跑的源程序，可以输入需要处理的图像，利用OpenCV生成test与test_g，当在板子上运行时，需要将测试的这两个文件拷贝到板子上去。