1. c++使用onnxruntime进行推理
link
code in git
#include <opencv2/core.hpp>
#include <opencv2/imgcodecs.hpp>
#include <opencv2/opencv.hpp>
#include <opencv2/highgui.hpp>
#include <opencv2/core/core.hpp>
#include <opencv2/imgproc/imgproc_c.h>
#include <opencv2/dnn.hpp>
#include <iostream>
#include <onnxruntime_cxx_api.h>
#include <assert.h>
#include <vector>
#include <fstream>
using namespace cv; //当定义这一行后,cv::imread可以直接写成imread
using namespace std;
using namespace Ort;
using namespace cv::dnn;
String labels_txt_file = "F:\\Pycharm\\PyCharm_Study\\Others\\c++_learning\\C++_Master\\Onnx\\classification\\classification_classes_ILSVRC2012.txt";
vector<String> readClassNames(); // string对象作为vector对象
// 图像处理 标准化处理
void PreProcess(const Mat& image, Mat& image_blob)
{
Mat input;
image.copyTo(input);
//数据处理 标准化
std::vector<Mat> channels, channel_p;
split(input, channels);
Mat R, G, B;
B = channels.at(0);
G = channels.at(1);
R = channels.at(2);
B = (B / 255. - 0.406) / 0.225;
G = (G / 255. - 0.456) / 0.224;
R = (R / 255. - 0.485) / 0.229;
channel_p.push_back(R);
channel_p.push_back(G);
channel_p.push_back(B);
Mat outt;
merge(channel_p, outt);
image_blob = outt;
}
// 读取txt文件
std::vector<String> readClassNames()
{
std::vector<String> classNames;
std::ifstream fp(labels_txt_file);
if (!fp.is_open())
{
printf("could not open file...\n");
exit(-1);
}
std::string name;
while (!fp.eof())
{
std::getline(fp, name);
if (name.length())
classNames.push_back(name);
}
fp.close();
return classNames;
}
int main() // 返回值为整型带参的main函数. 函数体内使用或不使用argc和argv都可
{
//environment (设置为VERBOSE(ORT_LOGGING_LEVEL_VERBOSE)时,方便控制台输出时看到是使用了cpu还是gpu执行)
Ort::Env env(ORT_LOGGING_LEVEL_WARNING, "OnnxModel");
Ort::SessionOptions session_options;
// 使用1个线程执行op,若想提升速度,增加线程数
session_options.SetIntraOpNumThreads(1);
CUDA加速开启(由于onnxruntime的版本太高,无cuda_provider_factory.h的头文件,加速可以使用onnxruntime V1.8的版本)
//OrtSessionOptionsAppendExecutionProvider_CUDA(session_options, 0);
// ORT_ENABLE_ALL: 启用所有可能的优化
session_options.SetGraphOptimizationLevel(GraphOptimizationLevel::ORT_ENABLE_ALL);
//load model and creat session
#ifdef _WIN32
const wchar_t* model_path = L"F:\\Pycharm\\PyCharm_Study\\Others\\c++_learning\\C++_Master\\Onnx\\classification\\vgg16.onnx";
#else
const char* model_path = "F:\\Pycharm\\PyCharm_Study\\Others\\c++_learning\\C++_Master\\Onnx\\classification\\vgg16.onnx";
#endif
printf("Using Onnxruntime C++ API\n");
Ort::Session session(env, model_path, session_options);
// print model input layer (node names, types, shape etc.)
Ort::AllocatorWithDefaultOptions allocator;
//model info
// 获得模型又多少个输入和输出,一般是指对应网络层的数目
// 一般输入只有图像的话input_nodes为1
size_t num_input_nodes = session.GetInputCount();
// 如果是多输出网络,就会是对应输出的数目
size_t num_output_nodes = session.GetOutputCount();
printf("Number of inputs = %zu\n", num_input_nodes);
printf("Number of output = %zu\n", num_output_nodes);
//获取输入name
const char* input_name = session.GetInputName(0, allocator);
std::cout << "input_name:" << input_name << std::endl;
//获取输出name
const char* output_name = session.GetOutputName(0, allocator);
std::cout << "output_name: " << output_name << std::endl;
// 自动获取维度数量
auto input_dims = session.GetInputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
auto output_dims = session.GetOutputTypeInfo(0).GetTensorTypeAndShapeInfo().GetShape();
std::cout << "input_dims:" << input_dims[0] << std::endl;
std::cout << "output_dims:" << output_dims[0] << std::endl;
std::vector<const char*> input_names{ input_name };
std::vector<const char*> output_names = { output_name };
std::vector<const char*> input_node_names = { "input.1" };
std::vector<const char*> output_node_names = { "70"};
//加载图片
Mat img = imread("F:\\Pycharm\\PyCharm_Study\\Others\\c++_learning\\C++_Master\\Onnx\\classification\\dog.jpg");
Mat det1, det2;
resize(img, det1, Size(256, 256), INTER_AREA);
det1.convertTo(det1, CV_32FC3);
PreProcess(det1, det2); //标准化处理
Mat blob = dnn::blobFromImage(det2, 1., Size(224, 224), Scalar(0, 0, 0), false, true);
printf("Load success!\n");
clock_t startTime, endTime;
//创建输入tensor
auto memory_info = Ort::MemoryInfo::CreateCpu(OrtAllocatorType::OrtArenaAllocator, OrtMemType::OrtMemTypeDefault);
std::vector<Ort::Value> input_tensors;
input_tensors.emplace_back(Ort::Value::CreateTensor<float>(memory_info, blob.ptr<float>(), blob.total(), input_dims.data(), input_dims.size()));
/*cout << int(input_dims.size()) << endl;*/
startTime = clock();
推理(score model & input tensor, get back output tensor)
auto output_tensors = session.Run(Ort::RunOptions{ nullptr }, input_node_names.data(), input_tensors.data(), input_names.size(), output_node_names.data(), output_node_names.size());
endTime = clock();
assert(output_tensors.size() == 1 && output_tensors.front().IsTensor());
//除了第一个节点外,其他参数与原网络对应不上程序就会无法执行
//第二个参数代表输入节点的名称集合
//第四个参数1代表输入层的数目
//第五个参数代表输出节点的名称集合
//最后一个参数代表输出节点的数目
获取输出(Get pointer to output tensor float values)
float* floatarr = output_tensors[0].GetTensorMutableData<float>(); // 也可以使用output_tensors.front(); 获取list中的第一个元素变量 list.pop_front(); 删除list中的第一个位置的元素
// 得到最可能分类输出
Mat newarr = Mat_<double>(1, 1000); //定义一个1*1000的矩阵
for (int i = 0; i < newarr.rows; i++)
{
for (int j = 0; j < newarr.cols; j++) //矩阵列数循环
{
newarr.at<double>(i, j) = floatarr[j];
}
}
/*cout << newarr.size() << endl;*/
vector<String> labels = readClassNames();
for (int n = 0; n < newarr.rows; n++) {
Point classNumber;
double classProb;
Mat probMat = newarr(Rect(0, n, 1000, 1)).clone();
Mat result = probMat.reshape(1, 1);
minMaxLoc(result, NULL, &classProb, NULL, &classNumber);
int classidx = classNumber.x;
printf("\n current image classification : %s, possible : %.2f\n", labels.at(classidx).c_str(), classProb);
// 显示文本
putText(img, labels.at(classidx), Point(10, 20), FONT_HERSHEY_SIMPLEX, 0.6, Scalar(0, 0, 255), 1, 1);
imshow("Image Classification", img);
waitKey(0);
}
计算运行时间
std::cout << "The run time is:" << (double)(endTime - startTime) / CLOCKS_PER_SEC << "s" << std::endl;
printf("Done!\n");
system("pause");
return 0;
}