用c++ opencv dnn调用keypointrcnn_resnet50_fpn.onnx模型写一个姿态估计示例，同时给出将keypointrcnn_resnet50_fpn转为onnx模型的示例

姿态估计示例：

#include <opencv2/opencv.hpp>
#include <opencv2/dnn.hpp>
#include <iostream>

using namespace cv;
using namespace cv::dnn;
using namespace std;

int main()
{
    // 加载模型
    Net net = readNetFromONNX("keypointrcnn_resnet50_fpn.onnx");

    // 打开摄像头
    VideoCapture cap(0);
    if (!cap.isOpened())
    {
        cout << "Error opening video stream" << endl;
        return -1;
    }

    while (true)
    {
        Mat frame;
        cap.read(frame);

        // 对图像进行预处理
        Mat inputBlob = blobFromImage(frame, 1.0 / 255, Size(1024, 1024), Scalar(0, 0, 0), true, false);
        net.setInput(inputBlob);

        // 进行推理
        vector<Mat> outs;
        net.forward(outs);

        // 获取关键点和边框坐标
        Mat detection = outs[0];
        Mat keypoints = outs[1];

        // 绘制关键点和边框
        for (int i = 0; i < detection.rows; i++)
        {
            float confidence = detection.at<float>(i, 2);
            if (confidence > 0.7)
            {
                int x1 = detection.at<float>(i, 3) * frame.cols;
                int y1 = detection.at<float>(i, 4) * frame.rows;
                int x2 = detection.at<float>(i, 5) * frame.cols;
                int y2 = detection.at<float>(i, 6) * frame.rows;
                rectangle(frame, Point(x1, y1), Point(x2, y2), Scalar(0, 255, 0), 2);

                for (int j = 0; j < keypoints.cols; j += 3)
                {
                    int x = keypoints.at<float>(i, j) * frame.cols;
                    int y = keypoints.at<float>(i, j + 1) * frame.rows;
                    float score = keypoints.at<float>(i, j + 2);
                    if (score > 0.3)
                        circle(frame, Point(x, y), 3, Scalar(0, 0, 255), -1);
                }
            }
        }

        // 显示结果
        imshow("Pose Estimation", frame);

        // 按Esc键退出
        if (waitKey(1) == 27)
            break;
    }

    cap.release();
    destroyAllWindows();

    return 0;
}

将keypointrcnn_resnet50_fpn转为onnx模型的示例：

import torch
import torchvision

# 加载模型
model = torchvision.models.detection.keypointrcnn_resnet50_fpn(pretrained=True)

# 设置模型为评估模式
model.eval()

# 导出模型为ONNX格式
input_shape = (3, 800, 800)
x = torch.randn(*input_shape)
torch.onnx.export(model, x, "keypointrcnn_resnet50_fpn.onnx", input_names=["input"], output_names=["detection", "keypoints"], opset_version=11)