mardi 1 août 2023

Errors when compiling OpenCV with GPU Support in Visual Studio C++ Windows10

I am attempting to install OpenCV with CUDA support on a new machine.

I expected the ALL_BUILD target to complete with no errors, allowing me to build the INSTALL target so I could use these modules in a project imported from another computer.

Instead, the ALL_BUILD target had 363 errors and it was not possible to continue building.

OS: Windows 10 Enterprise CPU: i7-13800H GPU: NVIDIA RTX A1000 6GB Laptop GPU C++ Standard: 11 OpenCV version: 4.5.5 OpenCV contrib version: 4.5.5 Visual Studio Version: Visual Studio 17 2022 CUDA version: 12.2 CUDNN version: 8.9.3

WITH_CUDA set to true. OPENCV_DNN_CUDA set to true. ENABLE_FAST_MATH set to true. CUDA_FAST_MATH set to true. CUDA_ARCH_BIN set to 8.6.

CUDA and CUDNN show up on Cmake GUI. When I build the ALL_BUILD Cmake target in Visual Studio, I get a total of 363 errors.

Most of these errors fall under 2 categories.

The first seems to be syntax errors in the common.hpp file. Here is a screenshot of the first of these errors: common.hpp errors

The second is a set of linking errors when attempting to compile OpenCV as shown here: LINK1181 input file errors

There are some other errors that occur as well. I have attached images of them as I am not sure which is important. MSB8066 Cuda Filters error texture is not a template no instance of overloaded function in opencv_xfeatures2d

Any advice on what to fix would be much appreciated.

Aucun commentaire:

Enregistrer un commentaire