OpenCV4实现YoloV3算法

 1、配置：在win10系统下，仅使用CPU(也可以使用GPU，根据电脑配置决定)，通过VS2015和OpenCV4.2.0编译实现。不过在Linux系统下编译，同时使用GPU加速最好。

2、代码中的配置文件，权重文件，以及数据下载：

配置文件+数据：https://github.com/PanJinquan/opencv-learning-tutorials

配置文件+权重文件：https://pjreddie.com/darknet/yolo

3、代码实现，该代码是根据网上的代码做简单的整理

yolov3.h：

#pragma once

#include 
#include 
#include 

#include 
#include 
#include 

using namespace cv;
using namespace std;

class YOLOV3
{
public:
	YOLOV3(float confThreshold = 0.5, float nmsThreshold = 0.4, int inpWidth = 416, int inpHeight = 416);

	void detect_image(std::string image_path, std::string modelWeights, std::string modelConfiguration, std::string classesFile, std::string& outputFile);
	void detect_video(std::string video_path, std::string modelWeights, std::string modelConfiguration, std::string classesFile, std::string& outputFile);

	// Remove the bounding boxes with low confidence using non-maxima suppression
	void postprocess(cv::Mat& frame, const std::vector& outs);

	// Get the names of the output layers
	std::vector getOutputsNames(const cv::dnn::Net& net);

	// Draw the predicted bounding box
	void drawPred(int classId, float conf, int left, int top, int right, int bottom, cv::Mat& frame);

private:

	// Initialize the parameters
	float mfConfThreshold;          // Confidence threshold
	float mfNmsThreshold;           // Non-maximum suppression threshold
	int mInpWidth;                  // Width of network's input image
	int mInpHeight;                 // Height of network's input image

	std::vector vClassIds;     // The index corresponding to the category name
	std::vector vClasses;   // Classification name of a category
	std::vector vConfidences;// Maximum confidence greater than confidence threshold
	std::vector vBoxes;   // Various category boxes
	std::vector vIndices;      // Candidate box index after non-maximum suppression
};

yolov3.cpp：

#include "yolov3.h"

YOLOV3::YOLOV3(float confThreshold, float nmsThreshold, int inpWidth, int inpHeight)
{
	mfConfThreshold = confThreshold;
	mfNmsThreshold = nmsThreshold;

	mInpWidth = inpWidth;
	mInpHeight = inpHeight;
}

void YOLOV3::detect_image(string image_path, string modelWeights, string modelConfiguration, string classesFile, std::string& outputFile)
{
	// Load names of vClasses
	ifstream ifs(classesFile.c_str());
	std::string line;
	while (getline(ifs, line)) vClasses.push_back(line);

	// Load the network
	dnn::Net net = dnn::readNetFromDarknet(modelConfiguration, modelWeights);
	net.setPreferableBackend(dnn::DNN_BACKEND_OPENCV);
	net.setPreferableTarget(dnn::DNN_TARGET_CPU);

	// Create a window
	static const string kWinName = "Deep learning object detection in OpenCV";
	namedWindow(kWinName, WINDOW_NORMAL);

	// Create a 4D blob from a frame.
	cv::Mat blob;
	cv::Mat frame = cv::imread(image_path);

	// Scale transformation, scaling, subtracting mean, channel transformation
	dnn::blobFromImage(frame, blob, 1 / 255.0, cv::Size(mInpWidth, mInpHeight), Scalar(0, 0, 0), true, false);

	// Sets the input to the network
	net.setInput(blob);

	// Runs the forward pass to get output of the output layers
	vector outs;
	net.forward(outs, getOutputsNames(net));

	// Remove the bounding boxes with low confidence
	postprocess(frame, outs);

	// Put efficiency information. The function getPerfProfile returns the overall time for inference(t) and the timings for each of the layers(in layersTimes)
	std::vector layersTimes;
	double freq = getTickFrequency() / 1000;
	double t = net.getPerfProfile(layersTimes) / freq;
	std::string label = format("Inference time for a frame : %.2f ms", t);

	putText(frame, label, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 0, 255));

	// Write the frame with the detection boxes
	imshow(kWinName, frame);
	cv::imwrite(outputFile, frame);
}

void YOLOV3::detect_video(string video_path, string modelWeights, string modelConfiguration, string classesFile, std::string& outputFile)
{
	// Load names of vClasses
	ifstream ifs(classesFile.c_str());
	string line;
	while (getline(ifs, line)) vClasses.push_back(line);

	// Load the network
	dnn::Net net = dnn::readNetFromDarknet(modelConfiguration, modelWeights);
	net.setPreferableBackend(dnn::DNN_BACKEND_OPENCV);
	net.setPreferableTarget(dnn::DNN_TARGET_CPU);

	// Open a video file or an image file or a camera stream.
	VideoCapture cap;
	VideoWriter video;

	Mat frame, blob;

	try {
		// Open the video file
		ifstream ifile(video_path);
		if (!ifile) throw("error");
		cap.open(video_path);
	}
	catch (...) {
		cout << "Could not open the input image/video stream" << endl;
		return;
	}

	// Get the video writer initialized to save the output video
	video.open(outputFile, VideoWriter::fourcc('M', 'J', 'P', 'G'), 28, Size(cap.get(CAP_PROP_FRAME_WIDTH), cap.get(CAP_PROP_FRAME_HEIGHT)));

	// Create a window
	static const string kWinName = "Deep learning object detection in OpenCV";
	namedWindow(kWinName, WINDOW_NORMAL);

	// Process frames.
	while (waitKey(1) < 0)
	{
		// get frame from the video
		cap >> frame;

		// Stop the program if reached end of video
		if (frame.empty()) {
			cout << "Done processing !!!" << endl;
			cout << "Output file is stored as " << outputFile << endl;
			waitKey(3000);
			break;
		}

		// Create a 4D blob from a frame.
		dnn::blobFromImage(frame, blob, 1 / 255.0, cv::Size(mInpWidth, mInpHeight), Scalar(0, 0, 0), true, false);

		//Sets the input to the network
		net.setInput(blob);

		// Runs the forward pass to get output of the output layers
		vector outs;
		net.forward(outs, getOutputsNames(net));

		// Remove the bounding boxes with low confidence
		postprocess(frame, outs);

		// Put efficiency information. The function getPerfProfile returns the overall time for inference(t) and the timings for each of the layers(in layersTimes)
		vector layersTimes;
		double freq = getTickFrequency() / 1000;
		double t = net.getPerfProfile(layersTimes) / freq;

		string label = format("Inference time for a frame : %.2f ms", t);

		cv::putText(frame, label, Point(0, 15), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 0, 255));

		// Write the frame with the detection boxes
		Mat detectedFrame;
		frame.convertTo(detectedFrame, CV_8U);
		video.write(detectedFrame);

		imshow(kWinName, frame);
	}

	cap.release();
	video.release();
}

// Scan through all the bounding boxes output from the network and keep only the
// ones with high confidence scores. Assign the box's class label as the class
// with the highest score for the box.
void YOLOV3::postprocess(cv::Mat& frame, const std::vector& outs)
{
	for (size_t i = 0; i < outs.size(); ++i)
	{
		float* data = (float*)outs[i].data;
		for (int j = 0; j < outs[i].rows; ++j, data += outs[i].cols)
		{
			cv::Point2i classIdPoint;
			double confidence;
			cv::Mat scores = outs[i].row(j).colRange(5, outs[i].cols);

			// Get the maximum score value in a matrix or vector and locate it
			minMaxLoc(scores, 0, &confidence, 0, &classIdPoint);

			if (confidence > mfConfThreshold)
			{
				// Get the parameters of the rectangular box. But what is the fifth data of "data"?
				int centerX = (int)(data[0] * frame.cols);
				int centerY = (int)(data[1] * frame.rows);
				int width = (int)(data[2] * frame.cols);
				int height = (int)(data[3] * frame.rows);

				// Should we consider the case if the rectangular frame crosses the image border? who??
				int left = centerX - width / 2;
				int top = centerY - height / 2;

				if (left < 0) left = 0;
				if (left < 0) top = 0;

				vClassIds.push_back(classIdPoint.x);
				vConfidences.push_back((float)confidence);
				vBoxes.push_back(Rect(left, top, width, height));
			}
		}
	}

	// Perform non maximum suppression to eliminate redundant overlapping boxes with lower confidences
	dnn::NMSBoxes(vBoxes, vConfidences, mfConfThreshold, mfNmsThreshold, vIndices);

	for (size_t i = 0; i < vIndices.size(); ++i)
	{
		int idx = vIndices[i];
		Rect box = vBoxes[idx];

		int right = box.x + box.width;
		int bottom = box.y + box.height;

		if (right > frame.cols) right = frame.cols;
		if (bottom > frame.rows) bottom = frame.rows;

		//Should we consider the case if the rectangular frame crosses the image border? who??
		drawPred(vClassIds[idx], vConfidences[idx], box.x, box.y, right, bottom, frame);
	}
}

// Draw the predicted bounding box
void YOLOV3::drawPred(int classId, float conf, int left, int top, int right, int bottom, Mat& frame)
{
	//Draw a rectangle displaying the bounding box
	rectangle(frame, Point(left, top), Point(right, bottom), Scalar(255, 178, 50), 3);

	//Get the label for the class name and its confidence
	string label = format("%.2f", conf);
	if (!vClasses.empty())
	{
		CV_Assert(classId < (int)vClasses.size());
		label = vClasses[classId] + ":" + label;
	}

	//Display the label at the top of the bounding box
	Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, NULL);
	top = max(top, labelSize.height);

	cv::rectangle(frame, Point(left, top - round(1.5*labelSize.height)), Point(left + round(1.5*labelSize.width), top + labelSize.height), Scalar(255, 255, 255), FILLED);
	cv::putText(frame, label, Point(left, top), FONT_HERSHEY_SIMPLEX, 0.75, Scalar(0, 0, 0), 1);
}

// Get the names of the output layers
std::vector YOLOV3::getOutputsNames(const cv::dnn::Net& net)
{
	static vector names;
	if (names.empty())
	{
		//Get the indices of the output layers, i.e. the layers with unconnected outputs
		vector outLayers = net.getUnconnectedOutLayers();

		//get the names of all the layers in the network
		vector layersNames = net.getLayerNames();

		// Get the names of the output layers in names
		names.resize(outLayers.size());
		for (size_t i = 0; i < outLayers.size(); ++i)
			names[i] = layersNames[outLayers[i] - 1];
	}
	return names;
}

main.cpp：

#include 
#include "yolov3.h"

int main(int argc, char** argv)
{
	// Give the configuration and weight files for the model
	string modelConfiguration =  "../data//models//yolov3//yolov3.cfg";
	string modelWeights =  "../data//models//yolov3//yolov3.weights";
	string classesFile = "../data//models//yolov3//coco.names";

	// Enter an image or video
	string image_path = "../data//images/dog.jpg";
	string video_path = "../data//images//run.mp4";

	// Output path settings
	std::string image_outputFile = "../result//yolov3.jpg";
	std::string video_outputFile = "../result//yolov3_out.avi";

	//Confidence threshold;Non-maximum suppression threshold;Width of network's input image;Height of network's input image
	YOLOV3 yolov3(0.5, 0.4, 416, 416);

	//yolov3.detect_image(image_path, modelWeights, modelConfiguration, classesFile, image_outputFile);
	yolov3.detect_video(video_path, modelWeights, modelConfiguration, classesFile, video_outputFile);

	cv::waitKey(0);
	return 0;
}

只需修改路径，即可实现视频流的检测。单张图像检测需要注释掉视频检测部分即可。

贴一张检测图片：

至于如何训练自己的数据集，在前面的博客有提到：https://blog.csdn.net/qq_38589460/article/details/85337032