


Set "python2_include_dir=C:/Python27/include" You can also explicitly specify the Python libraries for python 2 and python 3 in case the collector cannot find them automatically.
#Conda install opencv not working generator#
For Visual Studio 2017, the generator is written as “Visual StuWin64” and without the -A flag. Set "opencvExtraModules=opencv_contrib-4.1.1/modules" We open the console along the path C: \ OpenCV \ and register the variables. We will use the console version of cmake, since cmake-gui confuses the types of some variables (for example, OPENCV_PYTHON3_VERSION) and, as a result, generates files incorrectly. Next, we generate assembly files using cmake. git clone -b "4.1.1"Ĭreate the build / folder inside opencv-4.1.1. Next, download the opencv-4.1.1 and opencv-contrib-4.1.1 source archives to the desired location (in my case, this is C: \ OpenCV ). Also, if necessary, change the installation paths to your own.įirst you need to install the required software, and Visual Studio must be installed before CUDA:Īfter installing all the components, make sure that the paths for CMake, Visual Studio, Python, CUDA, CuDNN are written in the variables PATH, PYTHONPATH, CUDA_PATH and cudnn, respectively. Since the installation is done through console commands, you should carefully and accurately carry out all the steps. MS Visual Studio 2019 64-bit + CMake C ++ tools for Windows.The following tools were used in my assembly: CUDA 9 and below is supported by OpenCV 3.
#Conda install opencv not working windows 10#
Windows 10 + Visual Studio 2019 + Python 2/3 + CUDA 10.0 + GeForce GTX 1060Īttention! The build will not work for version OpenCV 4.0.1 and / or CUDA below version 10. Windows 8.1 + Visual Studio 2017 + Python 2/3 + CUDA 10.0 + GeForce 840m We tried to include here all the subtleties and nuances that you may encounter during the installation, and about which it is not written in the official manual. This article will focus on the assembly and installation of OpenCV 4 for C / C ++, Python 2 and Python 3 from source files with additional CUDA 10 modules on Windows.
