[人工智能]c++封装yolov4进行目标检测

yolo4是用c++写的，在工程中的部署特别方便。之前项目中使用yolov4，取得了不错的效果。在这里记录一下。
使用官方接口调用，我们首先得编译darknet动态库，下载yolov4源码

git clone https://github.com/AlexeyAB/darknet.git

编译yolov4为动态库

1. yolo_cpp_dll.sln，在darknet-master\build\darknet目录下面。
2. 设置x64和Lease，然后右击项目，生成。
   注意：这里应该配置好环境 CUDA 10.2以及 cuDNN，以及修改了yolo_cpp_dll.vcxproj中的cuda版本。
3. 然后在darknet-master\build\darknet\x64就会生成yolo_cpp_dll.dll文件。这个文件就是编译好的dll文件，可以直接使用。
   yolo_cpp_dll.dll的接口如下：

struct bbox_t {
    unsigned int x, y, w, h;    // (x,y) - top-left corner, (w, h) - width & height of bounded box
    float prob;                    // confidence - probability that the object was found correctly
    unsigned int obj_id;        // class of object - from range [0, classes-1]
    unsigned int track_id;        // tracking id for video (0 - untracked, 1 - inf - tracked object)
    unsigned int frames_counter;// counter of frames on which the object was detected
};

class Detector {
public:
        Detector(std::string cfg_filename, std::string weight_filename, int gpu_id = 0);
        ~Detector();

        std::vector<bbox_t> detect(std::string image_filename, float thresh = 0.2, bool use_mean = false);
        std::vector<bbox_t> detect(image_t img, float thresh = 0.2, bool use_mean = false);
        static image_t load_image(std::string image_filename);
        static void free_image(image_t m);

#ifdef OPENCV
        std::vector<bbox_t> detect(cv::Mat mat, float thresh = 0.2, bool use_mean = false);
        std::shared_ptr<image_t> mat_to_image_resize(cv::Mat mat) const;
#endif
};

关于这一部分有疑问的，可以参考官文档：https://github.com/AlexeyAB/darknet#yolo-v4-in-other-frameworks

检测步骤
检测部分需要用到刚刚编译的动态库，以及yolov4源码下的yolo_v2_class.hpp，需要将yolo_v2_class.hpp路径添加到包含目录中。

使用步骤大致分为以下几步
加载网络模型

string modelConfig = "yolov4/s_yolov4.cfg";
string modelWeights = "yolov4/yolov4.weights"";
string classesFile = "yolov4/s_coco.names";
Detector detector(modelConfig, modelWeights, 0);

调用 detector中的方法detect方法进行检测

std::vector<bbox_t> res = detector.detect(image);

image是Mat类型的

输出结果结构体bbox_t可以在yolo_v2_class.hpp看到，包括检测框左上角坐标，高和宽，置信度，类别等，有了它们画图也很简单了。

struct bbox_t {
    unsigned int x, y, w, h;       // (x,y) - top-left corner, (w, h) - width & height of bounded box
    float prob;                    // confidence - probability that the object was found correctly
    unsigned int obj_id;           // class of object - from range [0, classes-1]
    unsigned int track_id;         // tracking id for video (0 - untracked, 1 - inf - tracked object)
    unsigned int frames_counter;   // counter of frames on which the object was detected
    float x_3d, y_3d, z_3d;        // center of object (in Meters) if ZED 3D Camera is used
};

Drawer(image, res, classes);

void Drawer(Mat& frame, vector<bbox_t> outs, vector<string> classes)
{
	//获取所有最佳检测框信息
	for (int i = 0; i < outs.size(); i++)
	{
		DrawBoxes(frame, classes, outs[i].obj_id, outs[i].prob, outs[i].x, outs[i].y,
			outs[i].x + outs[i].w, outs[i].y + outs[i].h);
	}
}

void DrawBoxes(Mat& frame, vector<string> classes, int classId, float conf, int left, int top, int right, int bottom)
{
	//画检测框
	rectangle(frame, Point(left, top), Point(right, bottom), Scalar(255, 178, 50), 3);
	//该检测框对应的类别和置信度
	string label = format("%.2f", conf);
	if (!classes.empty())
	{
		CV_Assert(classId < (int)classes.size());
		label = classes[classId] + ":" + label;
	}
	//将标签显示在检测框顶部
	int baseLine;
	Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
	top = max(top, labelSize.height);
	//rectangle(frame, Point(left, top - round(1.5 * labelSize.height)), Point(left + round(1.5 * labelSize.width), top + baseLine), Scalar(255, 255, 255), FILLED);
	putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.75, Scalar(0, 0, 0), 1);
}

完整的代码如下：

#include <opencv2/opencv.hpp>
#include <yolo_v2_class.hpp>

using namespace std;
using namespace cv;

void DrawBoxes(Mat& frame, vector<string> classes, int classId, float conf, int left, int top, int right, int bottom)
{
	//画检测框
	rectangle(frame, Point(left, top), Point(right, bottom), Scalar(255, 178, 50), 3);
	//该检测框对应的类别和置信度
	string label = format("%.2f", conf);
	if (!classes.empty())
	{
		CV_Assert(classId < (int)classes.size());
		label = classes[classId] + ":" + label;
	}
	//将标签显示在检测框顶部
	int baseLine;
	Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
	top = max(top, labelSize.height);
	rectangle(frame, Point(left, top - round(1.5 * labelSize.height)), Point(left + round(1.5 * labelSize.width), top + baseLine), Scalar(255, 255, 255), FILLED);
	putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.75, Scalar(0, 0, 0), 1);
}

void Drawer(Mat& frame, vector<bbox_t> outs, vector<string> classes)
{
	//获取所有最佳检测框信息
	for (int i = 0; i < outs.size(); i++)
	{
		DrawBoxes(frame, classes, outs[i].obj_id, outs[i].prob, outs[i].x, outs[i].y,
			outs[i].x + outs[i].w, outs[i].y + outs[i].h);
	}
}

int main() {
	string classesFile = "yolov4/s_coco.names";
	vector<string> classes;
	ifstream ifs(classesFile.c_str());
	string line;
	while (getline(ifs, line)) classes.push_back(line);
	string modelConfig = "yolov4/yolov4.cfg";
	string modelWeights = "yolov4/yolov4.weights";
	Mat image = imread("dog.jpg");
	if (image.empty()) {
		cout << "the image is enpty" << endl;
	}
	else {
		Detector detector(modelConfig, modelWeights, 0);
		std::vector<bbox_t> res = detector.detect(image);
		Drawer(image, res, classes);
		imwrite("dog_result.jpg", image);
	}
	system("pause");
	return 0;
}

C#使用yolov4
由于在我的项目中，需要与C#的界面进行对接，并且需要实时获得检测结果，因此我想到将以上代码进一步封装成动态库。这样做的好处在于避免了C#直接调用yolo_cpp_dll.dll，方便在C++进行进一步的功能扩展。
在封装dll的过程中，不能直接对main方法中的内容进行封装。原因在于，如果对main中的内容都进行封装，每次检测图片都会经历一个模型加载的过程，会花费大量的时间，远远不能达到实时的效果。因此我对main中的内容进行了分解。
解决思路如下：
抽离加载模型的部分，返回模型的指针。

extern "C" __declspec(dllexport)
void* __stdcall model()
{
	string modelConfig = "yolov4/yolov4.cfg";
	string modelWeights = "yolov4/yolov4.weights";
	Detector* model = new Detector(modelConfig, modelWeights, 0);加载模型，只运行一遍
	return	(void*)model;
}

因为C#是可以接受C++的指针的，C#就可以将这个指针作为参数返回给C++的检测方法，对图片进行检测。
封装检测的部分代码如下：有两个参数，一个是图片指针，另一个是模型的指针。可以返回检测结果的指针，根据需要进行修改。

extern "C" __declspec(dllexport)
void* __stdcall Detect(const char* imagefile, void* p_model)
{
	Detector* model = (Detector*)p_model;
	string fileimage;
	fileimage = imagefile;
	Mat image = imread(fileimage);
	if (image.empty())
	{
		cout << "the image is enpty" << endl;
		return NULL;
	}
	std::vector<bbox_t> res = model->detect(image);
}

为了避免造成内存泄漏，C#需要调用C++的函数，将模型进行删除，删除模型的代码如下。

extern "C" __declspec(dllexport)
void __stdcall stop_model(void* p_model)
{
	Detector* model = (Detector*)p_model;
	delete(model);
	int* res_ar = (int*)res_arr;
}

最后：菜鸟一枚，欢迎各位大佬指正！！！！！！！！！