使用OpenCV、ONNXRuntime在c++平台部署YOLOV7实时目标检测的代码

好的，以下是一个简单的使用OpenCV和ONNXRuntime在C++平台部署YOLOv7实时目标检测的代码示例：

#include <iostream>
#include <chrono>
#include <opencv2/opencv.hpp>
#include <onnxruntime_cxx_api.h>

using namespace std;
using namespace cv;
using namespace std::chrono;
using namespace Ort;

int main() {
    // 加载模型
    Ort::Env env(ORT_LOGGING_LEVEL_WARNING, "test");
    Ort::SessionOptions session_options;
    OrtSession* session;
    Ort::AllocatorWithDefaultOptions allocator;
    const char* model_path = "yolov7.onnx";
    session = new Ort::Session(env, model_path, session_options);

    // 加载图像
    Mat frame = imread("test.jpg");
    if (frame.empty()) {
        cout << "读取图像失败！" << endl;
        return -1;
    }

    // 预处理图像
    Mat input_blob = blobFromImage(frame, 1 / 255.0, Size(416, 416), Scalar(0, 0, 0), true, false);
    Tensor input_tensor = Tensor(
        DataType::FLOAT,
        {1, 3, input_blob.rows, input_blob.cols},
        input_blob.data
    );

    // 运行模型
    vector<const char*> input_names = {"input_1"};
    vector<const char*> output_names = {"Identity_1", "Identity_2", "Identity_3"};
    vector<Ort::Value> input_tensors = {input_tensor};
    vector<Ort::Value> output_tensors = session->Run(Ort::RunOptions{}, input_names.data(), input_tensors.data(), input_names.size(), output_names.data(), output_names.size());

    // 后处理输出结果
    vector<Mat> detections;
    for (int i = 0; i < output_tensors.size(); i++) {
        float* output_data = output_tensors[i].GetTensorMutableData<float>();
        vector<int64_t> output_shape = output_tensors[i].GetTensorTypeAndShapeInfo().GetShape();
        Mat detection(output_shape[1], output_shape[0], CV_32F, output_data);
        detections.push_back(detection);
    }

    float confidence_threshold = 0.5;
    vector<int> class_ids;
    vector<float> confidences;
    vector<Rect> boxes;
    for (int i = 0; i < detections.size(); i++) {
        for (int j = 0; j < detections[i].rows; j++) {
            float* data = detections[i].ptr<float>(j);
            float confidence = data[4];
            if (confidence > confidence_threshold) {
                int class_id = j;
                float left = data[0] * frame.cols;
                float top = data[1] * frame.rows;
                float right = data[2] * frame.cols;
                float bottom = data[3] * frame.rows;
                int width = right - left + 1;
                int height = bottom - top + 1;
                class_ids.push_back(class_id);
                confidences.push_back(confidence);
                boxes.push_back(Rect(left, top, width, height));
            }
        }
    }

    for (int i = 0; i < boxes.size(); i++) {
        Scalar color(0, 255, 0);
        rectangle(frame, boxes[i], color, 2);
        string label = format("%.2f", confidences[i]);
        int baseline;
        Size label_size = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseline);
        rectangle(frame, Point(boxes[i].x, boxes[i].y - label_size.height - baseline), Point(boxes[i].x + label_size.width, boxes[i].y), color, FILLED);
        putText(frame, label, Point(boxes[i].x, boxes[i].y - baseline), FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0, 0, 0));
    }

    // 显示结果
    namedWindow("YOLOv7", WINDOW_NORMAL);
    resizeWindow("YOLOv7", 800, 600);
    imshow("YOLOv7", frame);
    waitKey(0);

    // 释放资源
    delete session;
    return 0;
}

这段代码首先使用ONNXRuntime加载YOLOv7模型，然后利用OpenCV读取图像并进行预处理，接着调用ONNXRuntime运行模型，最后根据模型输出结果进行后处理并显示检测结果。需要注意的是，这段代码只是一个简单的示例，实际部署时需要根据具体的场景进行调整。