1 安装opencv版本一定要3.4.12，否则与libtorch共同编译产生错误

linux上安装库或者程序，一个是编译安装、一个是apt安装，这里选择固定版本用编译安装
https://github.com/opencv/opencv/tags?after=4.5.0
https://cmake.org/download/

1 先下载个cmake，我用clion要求cmake20一下才行，直接调用bin下的cmake到clion就行
2下载opencv3.4.12源码
3 mkdir build，先安装一些依赖项，网上有，但是有两个库安装冲突，可以试试apitude install
4 cmake-gui 配置一下，再配置一下，再generate
5 make -j16 #我是8核心，一开始用j32直接死机
6 sudo make install # 这里将build好的文件安装到/usr/local/include share之类的，但是调用链接还得加到cmakelist
7 卸载进入build文件夹，make uninstall

2 安装libtorch

1 我按pytorch官方示例的wget的cpu版本,加压就行，因为是直接编译好的库文件
https://pytorch.org/cppdocs/installing.html
2 用的代码是之前看到知乎上的分类推理resnet18.pt
https://link.zhihu.com/?target=https%3A//github.com/BIGBALLON/PyTorch-CPP
https://blog.csdn.net/weixin_44523062/article/details/120132110

将下面代码在clion中打开，配置cmakelist 和调试参数 restnet18.pt label.txt
生成可执行程序加入参数运行就 ok 啦

C++代码

// One-stop header.
#include <torch/script.h>

#include <cmath>
#include <iostream>
#include <memory>
#include <opencv2/highgui/highgui.hpp>
#include <opencv2/imgproc/imgproc.hpp>
#include <opencv2/opencv.hpp>
#include <string>
#include <vector>

#define kIMAGE_SIZE 224
#define kCHANNELS 3
#define kTOP_K 3

bool LoadImage(std::string file_name, cv::Mat &image) {
  image = cv::imread(file_name);  // CV_8UC3
  if (image.empty() || !image.data) {
    return false;
  }
  cv::namedWindow("d");
  cv::imshow("d",image);
  cv::waitKey();
  cv::destroyAllWindows();
  cv::cvtColor(image, image, cv::COLOR_BGR2RGB);
  std::cout << "== image size: " << image.size() << " ==" << std::endl;

  // scale image to fit
  cv::Size scale(kIMAGE_SIZE, kIMAGE_SIZE);
  cv::resize(image, image, scale);
  std::cout << "== simply resize: " << image.size() << " ==" << std::endl;

  // convert [unsigned int] to [float]
  image.convertTo(image, CV_32FC3, 1.0f / 255.0f);

  return true;
}

bool LoadImageNetLabel(std::string file_name,
                       std::vector<std::string> &labels) {
  std::ifstream ifs(file_name);
  if (!ifs) {
    return false;
  }
  std::string line;
  while (std::getline(ifs, line)) {
    labels.push_back(line);
  }
  return true;
}

int main(int argc, const char *argv[]) {
  if (argc != 3) {
    std::cerr << "Usage: classifier <path-to-exported-script-module> "
                 "<path-to-lable-file>"
              << std::endl;
    return -1;
  }

  torch::jit::script::Module module = torch::jit::load(argv[1]);
  std::cout << "== Switch to GPU mode" << std::endl;
  // to GPU
//    module.to(at::kCUDA);
    module.to(at::kCPU);

  std::cout << "== Model [" << argv[1] << "] loaded!\n";
  std::vector<std::string> labels;
  if (LoadImageNetLabel(argv[2], labels)) {
    std::cout << "== Label loaded! Let's try it\n";
  } else {
    std::cerr << "Please check your label file path." << std::endl;
    return -1;
  }

  std::string file_name = "";
  cv::Mat image;
  while (true) {
    std::cout << "== Input image path: [enter Q to exit]" << std::endl;
    std::cin >> file_name;
    if (file_name == "Q") {
      break;
    }
    if (LoadImage(file_name, image)) {
      auto input_tensor = torch::from_blob(
          image.data, {1, kIMAGE_SIZE, kIMAGE_SIZE, kCHANNELS});
      input_tensor = input_tensor.permute({0, 3, 1, 2});
      input_tensor[0][0] = input_tensor[0][0].sub_(0.485).div_(0.229);
      input_tensor[0][1] = input_tensor[0][1].sub_(0.456).div_(0.224);
      input_tensor[0][2] = input_tensor[0][2].sub_(0.406).div_(0.225);

      // to GPU
//      input_tensor = input_tensor.to(at::kCUDA);
      input_tensor = input_tensor.to(at::kCPU);

      torch::Tensor out_tensor = module.forward({input_tensor}).toTensor();

      auto results = out_tensor.sort(-1, true);
      auto softmaxs = std::get<0>(results)[0].softmax(0);
      auto indexs = std::get<1>(results)[0];

      for (int i = 0; i < kTOP_K; ++i) {
        auto idx = indexs[i].item<int>();
        std::cout << "    ============= Top-" << i + 1
                  << " =============" << std::endl;
        std::cout << "    Label:  " << labels[idx] << std::endl;
        std::cout << "    With Probability:  "
                  << softmaxs[i].item<float>() * 100.0f << "%" << std::endl;
      }

    } else {
      std::cout << "Can't load the image, please check your path." << std::endl;
    }
  }
  return 0;
}

cmakelists

cmake_minimum_required(VERSION 3.20)
project(testlibtorch)

set(CMAKE_CXX_STANDARD 14)
set(Torch_DIR /home2/libtorch_apk/libtorch/share/cmake/Torch)

# 寻找OpenCV库
find_package( OpenCV 3 REQUIRED )
# 添加头文件
include_directories( ${OpenCV_INCLUDE_DIRS} )

find_package(Torch REQUIRED)
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${TORCH_CXX_FLAGS}")

add_executable(testlibtorch
        prediction.cpp)
target_link_libraries(testlibtorch "${TORCH_LIBRARIES}")
target_link_libraries( testlibtorch ${OpenCV_LIBS} )

set_property(TARGET testlibtorch PROPERTY CXX_STANDARD 14)

python导出模型程序

import torch
import torchvision

model = torchvision.models.resnet18(pretrained=True)

# Don't forget change model to eval mode
model.eval()
example = torch.rand(1, 3, 224, 224)
traced_script_module = torch.jit.trace(model, example)
traced_script_module.save("resnet18.pt")