能不能给我一个opencv的c++的yolo代码

以下是一个使用OpenCV C++封装的YOLOv4检测器的代码示例：

#include <iostream>
#include <fstream>
#include <sstream>
#include <opencv2/opencv.hpp>

class YOLO {
 public:
  YOLO(const std::string& cfg_path, const std::string& weights_path, float confidence_threshold = 0.5, float nms_threshold = 0.5, int input_width = 416, int input_height = 416, const std::vector<std::string>& class_names = {})
      : confidence_threshold_(confidence_threshold), nms_threshold_(nms_threshold), input_width_(input_width), input_height_(input_height), class_names_(class_names) {
    net_ = cv::dnn::readNetFromDarknet(cfg_path, weights_path);
    net_.setPreferableBackend(cv::dnn::DNN_BACKEND_CUDA);
    net_.setPreferableTarget(cv::dnn::DNN_TARGET_CUDA);
    output_names_ = net_.getUnconnectedOutLayersNames();
  }

  std::vector<cv::Rect> detect(const cv::Mat& image) {
    cv::Mat blob = cv::dnn::blobFromImage(image, 1 / 255.0, cv::Size(input_width_, input_height_), cv::Scalar(0, 0, 0), true, false);
    net_.setInput(blob);
    std::vector<cv::Mat> outputs;
    net_.forward(outputs, output_names_);

    std::vector<cv::Rect> boxes;
    std::vector<int> class_ids;
    std::vector<float> confidences;
    for (const auto& output : outputs) {
      cv::Mat detection = output.reshape(1, output.total() / 7);
      for (int i = 0; i < detection.rows; ++i) {
        cv::Mat row = detection.row(i);
        cv::Mat scores = row.colRange(5, detection.cols);
        cv::Point class_id;
        double confidence;
        cv::minMaxLoc(scores, nullptr, &confidence, nullptr, &class_id);
        if (confidence > confidence_threshold_) {
          int x_center = static_cast<int>(row.at<float>(0) * image.cols);
          int y_center = static_cast<int>(row.at<float>(1) * image.rows);
          int width = static_cast<int>(row.at<float>(2) * image.cols);
          int height = static_cast<int>(row.at<float>(3) * image.rows);
          int x_left = x_center - width / 2;
          int y_top = y_center - height / 2;
          class_ids.push_back(class_id.x);
          confidences.push_back(confidence);
          boxes.emplace_back(x_left, y_top, width, height);
        }
      }
    }

    std::vector<int> indices;
    cv::dnn::NMSBoxes(boxes, confidences, confidence_threshold_, nms_threshold_, indices);

    std::vector<cv::Rect> result;
    for (int i : indices) {
      result.push_back(boxes[i]);
    }
    return result;
  }

 private:
  cv::dnn::Net net_;
  float confidence_threshold_;
  float nms_threshold_;
  int input_width_;
  int input_height_;
  std::vector<std::string> class_names_;
  std::vector<std::string> output_names_;
};

int main() {
  std::string cfg_path = "/path/to/yolov4.cfg";
  std::string weights_path = "/path/to/yolov4.weights";
  std::vector<std::string> class_names = {"person", "car", "bicycle"};

  YOLO detector(cfg_path, weights_path, 0.5, 0.5, 416, 416, class_names);

  cv::Mat image = cv::imread("/path/to/image.jpg");
  std::vector<cv::Rect> boxes = detector.detect(image);
  for (const auto& box : boxes) {
    cv::rectangle(image, box, cv::Scalar(0, 0, 255), 2);
  }
  cv::imshow("Detection Result", image);
  cv::waitKey(0);

  return 0;
}

在此示例中，YOLO类的构造函数接受YOLOv4的配置文件路径和权重文件路径，并可选地传递置信度阈值、NMS阈值、输入图像的宽度和高度以及类别名称列表。如果不传递类别名称列表，则默认使用COCO数据集的80个类别名称。

YOLO类的detect方法接受一个OpenCV Mat对象作为输入，并返回一个包含检测结果边界框的矩形向量。在主函数中，YOLO检测器创建并使用detect方法来检测输入图像中的目标，并在原始图像上绘制检测结果的边界框。