windows环境下MFC调用libtorch

1.新建MFC应用程序，设置为release  x64运行模式，并设置为realease可断点调试模式（自行网上查询），其中SDL检查设置为否；

2.添加MFC程序所需环境变量：

vc++--包含目录：

E:\Software\OpenCV_2.4.13.5\opencv\build\include
E:\Software\OpenCV_2.4.13.5\opencv\build\include\opencv
E:\Software\OpenCV_2.4.13.5\opencv\build\include\opencv2
C:\Users\lvnengbin\Downloads\libtorch-win-shared-with-deps-1.1.0\libtorch\include
C:\Users\lvnengbin\Downloads\libtorch-win-shared-with-deps-1.1.0\libtorch\include\torch\csrc\api\include
C:\Program Files\NVIDIA Corporation\NvToolsExt\include
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.2\include

vc++--库目录：

E:\Software\OpenCV_2.4.13.5\opencv\build\x64\vc14\lib
C:\Users\lvnengbin\Downloads\libtorch-win-shared-with-deps-1.1.0\libtorch\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.2\lib\x64
C:\Program Files\NVIDIA Corporation\NvToolsExt\lib\x64

链接器--输入--附加依赖项：

torch.lib
c10.lib
caffe2.lib
nvToolsExt64_1.lib
cudart.lib
c10_cuda.lib
opencv_videostab2413.lib
opencv_ts2413.lib
opencv_superres2413.lib
opencv_stitching2413.lib
opencv_contrib2413.lib
caffe2_gpu.lib
cufft.lib
curand.lib
cudnn.lib
cublas.lib
opencv_nonfree2413.lib
opencv_ocl2413.lib
opencv_gpu2413.lib
opencv_photo2413.lib
opencv_objdetect2413.lib
opencv_legacy2413.lib
opencv_video2413.lib
opencv_ml2413.lib
opencv_calib3d2413.lib
opencv_features2d2413.lib
opencv_highgui2413.lib
opencv_imgproc2413.lib
opencv_flann2413.lib
opencv_core2413.lib
kernel32.lib
user32.lib
gdi32.lib
winspool.lib
shell32.lib
ole32.lib
oleaut32.lib
uuid.lib
comdlg32.lib
advapi32.lib
libprotobuf.lib

程序头文件：

#include "stdafx.h"

// 
// #include <iostream>  
// #include <memory>
#include "MFCLibtorchTest.h"
#include "MFCLibtorchTestDlg.h"
#include "afxdialogex.h"
#include<algorithm>

#undef UNICODE   //这个一定要加，否则会编译错误

//#include <torch/script.h> // 引入libtorch头文件
//#include "torch/torch.h"
#include <opencv2/core/core.hpp>  
#include <opencv2/highgui/highgui_c.h>
#include <opencv2/highgui/highgui.hpp>  
#include "opencv2/imgproc/imgproc.hpp"


#include <opencv2/opencv.hpp>
#include <string>
#include <vector>
#include <memory>
#include <torch/torch.h>
#include <torch/script.h> // One-stop header.

using namespace cv;

需要特别注意的是后面的头文件顺序很重要。通过测试发现，需要将torch相关头文件放在opencv相关库文件后面，就可以编译通过，可能是因为两者有部分编译是重复的吧。