Skip to content

BuildOpenCV4OpenVINO

Aleksandr Voron edited this page Sep 23, 2022 · 23 revisions
Clone this wiki locally

OpenCV usage with OpenVINO

Contents

Introduction

OpenVINO does not provide custom OpenCV drop since 2022.1.1 release. If OpenVINO user needs OpenCV functionality there are 2 approaches how to get it:

  1. Install prebuilt OpenCV from another sources (system repositories, pip, conda, homebrew). It is easy to follow this approach, however it has several disadvantages:
    • OpenCV version is out-of-date
    • OpenCV does not contain G-API module (e.g. some OMZ demos use G-API functionality)
    • OpenCV does not use available CPU instructions since it has build to cover wide range of hardware
    • OpenCV does not support Intel TBB, Intel Media SDK
    • OpenCV DNN module can not use OpenVINO as computational backend
  2. Build OpenCV from source code against specific version of OpenVINO. This approach solves the issues mentioned above.

Approach #1: Install prebuilt OpenCV

Installing on Ubuntu

OpenCV is available from the default Ubuntu repositories:

sudo apt update
sudo apt install libopencv-dev python3-opencv

Installing on RHEL

OpenCV is available from the default RHEL repositories.

  1. Enable CodeReady Linux Builder Repository:

subscription-manager repos --enable codeready-builder-for-rhel-8-x86_64-rpms

  1. Install OpenCV:

sudo yum update
sudo yum install opencv opencv-devel opencv-python

Installing on Windows

  1. Download OpenCV windows package on the official OpenCV releases page Windows archive contains VC postfix, e.g.: opencv-4.6.0-vc14_vc15.exe
  2. Extract the archive
  3. Setup environment variables by running the script opencv\build\setup_vars_opencv4.cmd

Installing on macOS

OpenCV could be installed via Homebrew:

brew install opencv

Approach #2: Build OpenCV against specific version of OpenVINO

NOTE: The instruction below describes how to build OpenCV with the same compilation flags and dependencies as were used for OpenVINO OpenCV build before 2022.1.1 release.
If you would like to use a minimal set of compilation flags and dependencies to build OpenCV compatible with OpenVINO please follow another instruction.

Building on Ubuntu

Prerequisites

  1. Install OpenVINO according to the instruction
  2. Install the following packages:
The command

sudo apt-get install \
build-essential \
cmake \
ninja-build \
libgtk-3-dev \
libpng-dev \
libjpeg-dev \
libwebp-dev \
libtiff5-dev \
libopenexr-dev \
libopenblas-dev \
libx11-dev \
libavutil-dev \
libavcodec-dev \
libavformat-dev \
libswscale-dev \
libavresample-dev \
libtbb2 \
libssl-dev \
libva-dev \
libmfx-dev \
libgstreamer1.0-dev \
libgstreamer-plugins-base1.0-dev

Procedure

  1. Setup environment variables to detect OpenVINO:

source /opt/intel/openvino/bin/setupvars.sh

  1. Copy OpenCV repository:

git clone --recurse-submodules https://github.com/opencv/opencv.git

  1. Create build directory and enter into it:

mkdir ~/build-opencv && cd ~/build-opencv

  1. Compile and install OpenCV:
The command

cmake -G Ninja \
-D BUILD_INFO_SKIP_EXTRA_MODULES=ON \
-D BUILD_EXAMPLES=OFF \
-D BUILD_JASPER=OFF \
-D BUILD_JAVA=OFF \
-D BUILD_JPEG=ON \
-D BUILD_APPS_LIST=version \
-D BUILD_opencv_apps=ON \
-D BUILD_opencv_java=OFF \
-D BUILD_OPENEXR=OFF \
-D BUILD_PNG=ON \
-D BUILD_TBB=OFF \
-D BUILD_WEBP=OFF \
-D BUILD_ZLIB=ON \
-D WITH_1394=OFF \
-D WITH_CUDA=OFF \
-D WITH_EIGEN=OFF \
-D WITH_GPHOTO2=OFF \
-D WITH_GSTREAMER=ON \
-D OPENCV_GAPI_GSTREAMER=OFF \
-D WITH_GTK_2_X=OFF \
-D WITH_IPP=ON \
-D WITH_JASPER=OFF \
-D WITH_LAPACK=OFF \
-D WITH_MATLAB=OFF \
-D WITH_MFX=ON \
-D WITH_OPENCLAMDBLAS=OFF \
-D WITH_OPENCLAMDFFT=OFF \
-D WITH_OPENEXR=OFF \
-D WITH_OPENJPEG=OFF \
-D WITH_QUIRC=OFF \
-D WITH_TBB=OFF \
-D WITH_TIFF=OFF \
-D WITH_VTK=OFF \
-D WITH_WEBP=OFF \
-D CMAKE_USE_RELATIVE_PATHS=ON \
-D CMAKE_SKIP_INSTALL_RPATH=ON \
-D ENABLE_BUILD_HARDENING=ON \
-D ENABLE_CONFIG_VERIFICATION=ON \
-D ENABLE_PRECOMPILED_HEADERS=OFF \
-D ENABLE_CXX11=ON \
-D INSTALL_PDB=ON \
-D INSTALL_TESTS=ON \
-D INSTALL_C_EXAMPLES=ON \
-D INSTALL_PYTHON_EXAMPLES=ON \
-D CMAKE_INSTALL_PREFIX=install \
-D OPENCV_SKIP_PKGCONFIG_GENERATION=ON \
-D OPENCV_SKIP_PYTHON_LOADER=OFF \
-D OPENCV_SKIP_CMAKE_ROOT_CONFIG=ON \
-D OPENCV_GENERATE_SETUPVARS=OFF \
-D OPENCV_BIN_INSTALL_PATH=bin \
-D OPENCV_INCLUDE_INSTALL_PATH=include \
-D OPENCV_LIB_INSTALL_PATH=lib \
-D OPENCV_CONFIG_INSTALL_PATH=cmake \
-D OPENCV_3P_LIB_INSTALL_PATH=3rdparty \
-D OPENCV_SAMPLES_SRC_INSTALL_PATH=samples \
-D OPENCV_DOC_INSTALL_PATH=doc \
-D OPENCV_OTHER_INSTALL_PATH=etc \
-D OPENCV_LICENSES_INSTALL_PATH=etc/licenses \
-D OPENCV_INSTALL_FFMPEG_DOWNLOAD_SCRIPT=ON \
-D BUILD_opencv_world=OFF \
-D BUILD_opencv_python2=OFF \
-D BUILD_opencv_python3=ON \
-D PYTHON3_PACKAGES_PATH=install/python/python3 \
-D PYTHON3_LIMITED_API=ON \
-D HIGHGUI_PLUGIN_LIST=all \
-D OPENCV_PYTHON_INSTALL_PATH=python \
-D CPU_BASELINE=SSE4_2 \
-D OPENCV_IPP_GAUSSIAN_BLUR=ON \
-D WITH_OPENVINO=ON \
-D VIDEOIO_PLUGIN_LIST=ffmpeg,gstreamer,mfx \
-D CMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined \
-D CMAKE_BUILD_TYPE=Release <OpenCV_ROOT_REPO_DIRECTORY> && \
ninja && cmake --install .

OpenCV package is available at ~/build-opencv/install directory.

To compile application that uses OpenCV, the following environment variables should be specified:

export OpenCV_DIR="<OpenCV_INSTALL_DIR>/cmake"
export LD_LIBRARY_PATH="<OpenCV_INSTALL_DIR>/lib${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
export PYTHONPATH="<OpenCV_INSTALL_DIR>/python${PYTHONPATH:+:$PYTHONPATH}"

Building on RHEL

Prerequisites

  1. Install OpenVINO according to the instruction
  2. Install EPEL and RPM Fusion repositories using the following commands:

yum install https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm https://download1.rpmfusion.org/free/el/rpmfusion-free-release-8.noarch.rpm https://download1.rpmfusion.org/nonfree/el/rpmfusion-nonfree-release-8.noarch.rpm

  1. Install the following packages:
The command

sudo yum install \
cmake \
gtk3-devel \
libmfx-devel \
gstreamer1-devel \
gstreamer1-plugins-base-devel \
ffmpeg-devel

Procedure

  1. Setup environment variables to detect OpenVINO:

source /opt/intel/openvino/bin/setupvars.sh

  1. Copy OpenCV repository:

git clone --recurse-submodules https://github.com/opencv/opencv.git

  1. Create build directory and enter into it:

mkdir ~/build-opencv && cd ~/build-opencv

  1. Compile and install OpenCV:
The command

cmake \
-D BUILD_INFO_SKIP_EXTRA_MODULES=ON \
-D BUILD_EXAMPLES=OFF \
-D BUILD_JASPER=OFF \
-D BUILD_JAVA=OFF \
-D BUILD_JPEG=ON \
-D BUILD_APPS_LIST=version \
-D BUILD_opencv_apps=ON \
-D BUILD_opencv_java=OFF \
-D BUILD_OPENEXR=OFF \
-D BUILD_PNG=ON \
-D BUILD_TBB=OFF \
-D BUILD_WEBP=OFF \
-D BUILD_ZLIB=ON \
-D WITH_1394=OFF \
-D WITH_CUDA=OFF \
-D WITH_EIGEN=OFF \
-D WITH_GPHOTO2=OFF \
-D WITH_GSTREAMER=ON \
-D OPENCV_GAPI_GSTREAMER=OFF \
-D WITH_GTK_2_X=OFF \
-D WITH_IPP=ON \
-D WITH_JASPER=OFF \
-D WITH_LAPACK=OFF \
-D WITH_MATLAB=OFF \
-D WITH_MFX=OFF \
-D WITH_OPENCLAMDBLAS=OFF \
-D WITH_OPENCLAMDFFT=OFF \
-D WITH_OPENEXR=OFF \
-D WITH_OPENJPEG=OFF \
-D WITH_QUIRC=OFF \
-D WITH_TBB=OFF \
-D WITH_TIFF=OFF \
-D WITH_VTK=OFF \
-D WITH_WEBP=OFF \
-D CMAKE_USE_RELATIVE_PATHS=ON \
-D CMAKE_SKIP_INSTALL_RPATH=ON \
-D ENABLE_BUILD_HARDENING=ON \
-D ENABLE_CONFIG_VERIFICATION=ON \
-D ENABLE_PRECOMPILED_HEADERS=OFF \
-D ENABLE_CXX11=ON \
-D INSTALL_PDB=ON \
-D INSTALL_TESTS=ON \
-D INSTALL_C_EXAMPLES=ON \
-D INSTALL_PYTHON_EXAMPLES=ON \
-D CMAKE_INSTALL_PREFIX=install \
-D OPENCV_SKIP_PKGCONFIG_GENERATION=ON \
-D OPENCV_SKIP_PYTHON_LOADER=OFF \
-D OPENCV_SKIP_CMAKE_ROOT_CONFIG=ON \
-D OPENCV_GENERATE_SETUPVARS=OFF \
-D OPENCV_BIN_INSTALL_PATH=bin \
-D OPENCV_INCLUDE_INSTALL_PATH=include \
-D OPENCV_LIB_INSTALL_PATH=lib \
-D OPENCV_CONFIG_INSTALL_PATH=cmake \
-D OPENCV_3P_LIB_INSTALL_PATH=3rdparty \
-D OPENCV_SAMPLES_SRC_INSTALL_PATH=samples \
-D OPENCV_DOC_INSTALL_PATH=doc \
-D OPENCV_OTHER_INSTALL_PATH=etc \
-D OPENCV_LICENSES_INSTALL_PATH=etc/licenses \
-D OPENCV_INSTALL_FFMPEG_DOWNLOAD_SCRIPT=ON \
-D BUILD_opencv_world=OFF \
-D BUILD_opencv_python2=OFF \
-D BUILD_opencv_python3=ON \
-D PYTHON3_PACKAGES_PATH=install/python/python3 \
-D PYTHON3_LIMITED_API=ON \
-D HIGHGUI_PLUGIN_LIST=all \
-D OPENCV_PYTHON_INSTALL_PATH=python \
-D CPU_BASELINE=SSE4_2 \
-D OPENCV_IPP_GAUSSIAN_BLUR=ON \
-D WITH_OPENVINO=ON \
-D VIDEOIO_PLUGIN_LIST=ffmpeg,gstreamer \
-D CMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined \
-D CMAKE_BUILD_TYPE=Release <OpenCV_ROOT_REPO_DIRECTORY> && \
make && cmake -P cmake_install.cmake

OpenCV package is available at ~/build-opencv/install directory.

To compile application that uses OpenCV, the following environment variables should be specified:

export OpenCV_DIR="<OpenCV_INSTALL_DIR>/cmake"
export LD_LIBRARY_PATH="<OpenCV_INSTALL_DIR>/lib${LD_LIBRARY_PATH:+:$LD_LIBRARY_PATH}"
export PYTHONPATH="<OpenCV_INSTALL_DIR>/python${PYTHONPATH:+:$PYTHONPATH}"

Building on Windows

Prerequisites

  1. Install Microsoft Visual Studio
  2. Install cmake
  3. Install OpenVINO according to the instruction
  4. Install Intel® Media SDK for Windows

Procedure

  1. Setup environment variables to detect OpenVINO:

"C:\Program Files (x86)\IntelSWTools\openvino\bin\setupvars.bat"

  1. Copy OpenCV repository:

git clone --recurse-submodules https://github.com/opencv/opencv.git

  1. Create build directory and enter into it:

mkdir "build-opencv" && cd "build-opencv"

  1. Setup MSVC environment by running vcvars64.bat
  2. Compile and install OpenCV:
The command

cmake -G Ninja ^
-DBUILD_INFO_SKIP_EXTRA_MODULES=ON ^
-DBUILD_EXAMPLES=OFF ^
-DBUILD_JASPER=OFF ^
-DBUILD_JAVA=OFF ^
-DBUILD_JPEG=ON ^
-DBUILD_APPS_LIST=version ^
-DBUILD_opencv_apps=ON ^
-DBUILD_opencv_java=OFF ^
-DBUILD_OPENEXR=OFF ^
-DBUILD_PNG=ON ^
-DBUILD_TBB=OFF ^
-DBUILD_WEBP=OFF ^
-DBUILD_ZLIB=ON ^
-DWITH_1394=OFF ^
-DWITH_CUDA=OFF ^
-DWITH_EIGEN=OFF ^
-DWITH_GPHOTO2=OFF ^
-DWITH_GSTREAMER=OFF ^
-DOPENCV_GAPI_GSTREAMER=OFF ^
-DWITH_GTK_2_X=OFF ^
-DWITH_IPP=ON ^
-DWITH_JASPER=OFF ^
-DWITH_LAPACK=OFF ^
-DWITH_MATLAB=OFF ^
-DWITH_MFX=ON ^
-DWITH_OPENCLAMDBLAS=OFF ^
-DWITH_OPENCLAMDFFT=OFF ^
-DWITH_OPENEXR=OFF ^
-DWITH_OPENJPEG=OFF ^
-DWITH_QUIRC=OFF ^
-DWITH_TBB=OFF ^
-DWITH_TIFF=OFF ^
-DWITH_VTK=OFF ^
-DWITH_WEBP=OFF ^
-DCMAKE_USE_RELATIVE_PATHS=ON ^
-DCMAKE_SKIP_INSTALL_RPATH=ON ^
-DENABLE_BUILD_HARDENING=ON ^
-DENABLE_CONFIG_VERIFICATION=ON ^
-DENABLE_PRECOMPILED_HEADERS=OFF ^
-DENABLE_CXX11=ON ^
-DINSTALL_PDB=ON ^
-DINSTALL_TESTS=ON ^
-DINSTALL_C_EXAMPLES=ON ^
-DINSTALL_PYTHON_EXAMPLES=ON ^
-DCMAKE_INSTALL_PREFIX=install ^
-DOPENCV_SKIP_PKGCONFIG_GENERATION=ON ^
-DOPENCV_SKIP_PYTHON_LOADER=OFF ^
-DOPENCV_SKIP_CMAKE_ROOT_CONFIG=ON ^
-DOPENCV_GENERATE_SETUPVARS=OFF ^
-DOPENCV_BIN_INSTALL_PATH=bin ^
-DOPENCV_INCLUDE_INSTALL_PATH=include ^
-DOPENCV_LIB_INSTALL_PATH=lib ^
-DOPENCV_CONFIG_INSTALL_PATH=cmake ^
-DOPENCV_3P_LIB_INSTALL_PATH=3rdparty ^
-DOPENCV_SAMPLES_SRC_INSTALL_PATH=samples ^
-DOPENCV_DOC_INSTALL_PATH=doc ^
-DOPENCV_OTHER_INSTALL_PATH=etc ^
-DOPENCV_LICENSES_INSTALL_PATH=etc/licenses ^
-DOPENCV_INSTALL_FFMPEG_DOWNLOAD_SCRIPT=ON ^
-DBUILD_opencv_world=OFF ^
-DBUILD_opencv_python2=OFF ^
-DBUILD_opencv_python3=ON ^
-DPYTHON3_PACKAGES_PATH=install/python/python3 ^
-DPYTHON3_LIMITED_API=ON ^
-DOPENCV_PYTHON_INSTALL_PATH=python ^
-DCPU_BASELINE=SSE4_2 ^
-DOPENCV_IPP_GAUSSIAN_BLUR=ON ^
-DWITH_OPENVINO=ON ^
-DVIDEOIO_PLUGIN_LIST=mfx,msmf ^
-DCMAKE_BUILD_TYPE=Release <OpenCV_ROOT_REPO_DIRECTORY> &&
ninja &&
cmake --install .

OpenCV package is available at build-opencv/install directory.

To compile application that uses OpenCV, the following environment variables should be specified:

set "OpenCV_DIR=<OpenCV_INSTALL_DIR>\cmake"
set "PATH=<OpenCV_INSTALL_DIR>\bin;%PATH%"
set "PYTHONPATH=<OpenCV_INSTALL_DIR>\python;%PYTHONPATH%"

Building on macOS

Prerequisites

  1. Install OpenVINO according to the instruction
  2. Install the following packages:

brew install cmake \
ninja \
ffmpeg \
gstreamer \
gst-devtools
python \
numpy

Procedure

  1. Setup environment variables to detect OpenVINO:

source /opt/intel/openvino/bin/setupvars.sh

  1. Copy OpenCV repository:

git clone --recurse-submodules https://github.com/opencv/opencv.git

  1. Create build directory and enter into it:

mkdir "build-opencv" && cd "build-opencv"

  1. Compile and install OpenCV:
The command

cmake \
-G Ninja \
-DBUILD_INFO_SKIP_EXTRA_MODULES=ON \
-DBUILD_EXAMPLES=OFF \
-DBUILD_JASPER=OFF \
-DBUILD_JAVA=OFF \
-DBUILD_JPEG=ON \
-DBUILD_APPS_LIST=version \
-DBUILD_opencv_apps=ON \
-DBUILD_opencv_java=OFF \
-DBUILD_OPENEXR=OFF \
-DBUILD_PNG=ON \
-DBUILD_TBB=OFF \
-DBUILD_WEBP=OFF \
-DBUILD_ZLIB=ON \
-DWITH_1394=OFF \
-DWITH_CUDA=OFF \
-DWITH_EIGEN=OFF \
-DWITH_GPHOTO2=OFF \
-DWITH_GSTREAMER=ON \
-DOPENCV_GAPI_GSTREAMER=OFF \
-DWITH_GTK_2_X=OFF \
-DWITH_IPP=ON \
-DWITH_JASPER=OFF \
-DWITH_LAPACK=OFF \
-DWITH_MATLAB=OFF \
-DWITH_MFX=OFF \
-DWITH_OPENCLAMDBLAS=OFF \
-DWITH_OPENCLAMDFFT=OFF \
-DWITH_OPENEXR=OFF \
-DWITH_OPENJPEG=OFF \
-DWITH_QUIRC=OFF \
-DWITH_TBB=OFF \
-DWITH_TIFF=OFF \
-DWITH_VTK=OFF \
-DWITH_WEBP=OFF \
-DCMAKE_USE_RELATIVE_PATHS=ON \
-DCMAKE_SKIP_INSTALL_RPATH=ON \
-DENABLE_BUILD_HARDENING=ON \
-DENABLE_CONFIG_VERIFICATION=ON \
-DENABLE_PRECOMPILED_HEADERS=OFF \
-DENABLE_CXX11=ON \
-DINSTALL_PDB=ON \
-DINSTALL_TESTS=ON \
-DINSTALL_C_EXAMPLES=ON \
-DINSTALL_PYTHON_EXAMPLES=ON \
-DCMAKE_INSTALL_PREFIX=install \
-DOPENCV_SKIP_PKGCONFIG_GENERATION=ON \
-DOPENCV_SKIP_PYTHON_LOADER=OFF \
-DOPENCV_SKIP_CMAKE_ROOT_CONFIG=ON \
-DOPENCV_GENERATE_SETUPVARS=OFF \
-DOPENCV_BIN_INSTALL_PATH=bin \
-DOPENCV_INCLUDE_INSTALL_PATH=include \
-DOPENCV_LIB_INSTALL_PATH=lib \
-DOPENCV_CONFIG_INSTALL_PATH=cmake \
-DOPENCV_3P_LIB_INSTALL_PATH=3rdparty \
-DOPENCV_SAMPLES_SRC_INSTALL_PATH=samples \
-DOPENCV_DOC_INSTALL_PATH=doc \
-DOPENCV_OTHER_INSTALL_PATH=etc \
-DOPENCV_LICENSES_INSTALL_PATH=etc/licenses \
-DOPENCV_INSTALL_FFMPEG_DOWNLOAD_SCRIPT=ON \
-DBUILD_opencv_world=OFF \
-DBUILD_opencv_python2=OFF \
-DBUILD_opencv_python3=ON \
-DPYTHON3_PACKAGES_PATH=install/python/python3 \
-DPYTHON3_LIMITED_API=ON \
-DOPENCV_PYTHON_INSTALL_PATH=python \
-DCPU_BASELINE=SSE4_2 \
-DWITH_AVFOUNDATION=ON \
-DOPENCV_IPP_GAUSSIAN_BLUR=ON \
-DWITH_OPENVINO=ON \
-DVIDEOIO_PLUGIN_LIST=ffmpeg,gstreamer \
-DCMAKE_BUILD_TYPE=Release <OpenCV_ROOT_REPO_DIRECTORY> &&
ninja &&
cmake --install .

OpenCV package is available at build-opencv/install directory.

To compile application that uses OpenCV, the following environment variables should be specified:

export OpenCV_DIR="<OpenCV_INSTALL_DIR>/cmake"