vs2015+qt 调用cpu版本darknet

1.编译cpu版本darknet生成动态链接库

先将cpu版本的yolo_cpp_dll_no_gpu进行编译，在Release版本下编译生成yolo_cpp_dll.dll和yolo_cpp_dll_no_gpu.lib，并将yolo_cpp_dll.dll改名为yolo_cpp_dll_no_gpu.dll

2.新建一个qt工程，release文件下放入4个动态链接库dll文件，包括opencv_world340.dll，另外3个dll文件都在darknet-master\build\darknet\x64下找到

3.main.cpp所在目录下放入yolo_cpp_dll_no_gpu.lib文件，3个Yolo模型文件voc.names/cfg/weights，以及yolo_v2_class.hpp头文件，并在qt工程的头文件点击添加现有文件yolo_v2_class.hpp（yolo_v2_class.hpp在darknet-master\build\darknet\yolo_console_dll打开后可以看到，复制到记事本里）

4.加入代码

#include "imageProcess.h"
#include "stdafx.h"
#include <fstream>//忘记添加导致出现"不允许使用不完整的类型"错误

#define OPENCV   //define OPENCV必须要在include "yolo_v2_class.hpp" 之前
#include "yolo_v2_class.hpp" //引用动态链接库中的头文件
#pragma comment(lib, "yolo_cpp_dll_no_gpu.lib") //引入YOLO动态链接库

CImgProcess::CImgProcess()
{

}

CImgProcess::~CImgProcess()
{
    // Stop media playback
}

void draw_boxes(cv::Mat mat_img, std::vector<bbox_t> result_vec, std::vector<std::string> obj_names,
    int current_det_fps = -1, int current_cap_fps = -1)
{
    int const colors[6][3] = { { 1,0,1 },{ 0,0,1 },{ 0,1,1 },{ 0,1,0 },{ 1,1,0 },{ 1,0,0 } };

    for (auto &i : result_vec) {
        cv::Scalar color = obj_id_to_color(i.obj_id);
        cv::rectangle(mat_img, cv::Rect(i.x, i.y, i.w, i.h), color, 2);
        if (obj_names.size() > i.obj_id) {
            std::string obj_name = obj_names[i.obj_id];
            if (i.track_id > 0) obj_name += " - " + std::to_string(i.track_id);
            cv::Size const text_size = getTextSize(obj_name, cv::FONT_HERSHEY_COMPLEX_SMALL, 1.2, 2, 0);
            int const max_width = (text_size.width > i.w + 2) ? text_size.width : (i.w + 2);
            cv::rectangle(mat_img, cv::Point2f(std::max((int)i.x - 1, 0), std::max((int)i.y - 30, 0)),
                cv::Point2f(std::min((int)i.x + max_width, mat_img.cols - 1), std::min((int)i.y, mat_img.rows - 1)),
                color, CV_FILLED, 8, 0);
            putText(mat_img, obj_name, cv::Point2f(i.x, i.y - 10), cv::FONT_HERSHEY_COMPLEX_SMALL, 1.2, cv::Scalar(0, 0, 0), 2);
        }
    }
    if (current_det_fps >= 0 && current_cap_fps >= 0) {
        std::string fps_str = "FPS detection: " + std::to_string(current_det_fps) + "   FPS capture: " + std::to_string(current_cap_fps);
        putText(mat_img, fps_str, cv::Point2f(10, 20), cv::FONT_HERSHEY_COMPLEX_SMALL, 1.2, cv::Scalar(50, 255, 0), 2);
    }
}

std::vector<std::string> objects_names_from_file(std::string const filename)
{
    std::ifstream file(filename);
    std::vector<std::string> file_lines;
    if (!file.is_open()) return file_lines;
    for (std::string line; getline(file, line);) file_lines.push_back(line);
    std::cout << "object names loaded \n";
    return file_lines;
}

void CImgProcess::ImgProcess_second()//图像处理函数
{

    std::string names_file = "voc.names";
    std::string cfg_file = "yolov3-voc.cfg";
    std::string weights_file = "yolov3-voc_last.weights";

                                                  
    Detector detector(cfg_file, weights_file, 1); //初始化检测器

    std::vector<std::string> obj_names;
    std::ifstream ifs(names_file.c_str());
    std::string line;
    while (getline(ifs, line)) obj_names.push_back(line);

    double time0 = (double)(getTickCount());
    std::vector<bbox_t> result_vec = detector.detect(img);

    for (auto &i : result_vec)
    {
        cv::Scalar color = obj_id_to_color(i.obj_id);
        cv::rectangle(img, cv::Rect(i.x, i.y, i.w, i.h), color, 2);
    }
    totaltime = 1000 * ((double)getTickCount() - time0) / getTickFrequency();//1000 *总次数/一秒内重复的次数= 时间(ms)

}