Skip to content
This repository has been archived by the owner on Mar 10, 2024. It is now read-only.

Installing realsense opencv cuda

Jump to bottom
emcauliffe edited this page May 25, 2023 · 1 revision

Installing OpenCV and Intel RealSense with CUDA support

Building and installing OpenCV from source

The instructions below are for openCV 4.7.0 (latest at time of writing)

  1. Download the latest opencv source code release and opencv-contrib from the official git repos. Make sure both folders exist in the same parent directory.

    https://github.com/opencv/opencv/releases

    wget https://github.com/opencv/opencv/archive/refs/tags/4.7.0.zip

    unzip 4.7.0.zip

    rm 4.7.0.zip

    https://github.com/opencv/opencv_contrib/tags

    wget https://github.com/opencv/opencv_contrib/archive/refs/tags/4.7.0.zip

    unzip 4.7.0.zip

    rm 4.7.0.zip

  2. In the same parent directory, make the build repo for opencv

    mkdir opencv-4.7.0_build

    and enter the build directory

    cd opencv-4.7.0_build

  3. Run CMake to configure the OpenCV build

    cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D WITH_CUDA=ON -D CUDA_ARCH_BIN=6.2 -D CUDA_ARCH_PTX="" -D ENABLE_FAST_MATH=ON -D CUDA_FAST_MATH=ON -D WITH_CUBLAS=ON -D WITH_LIBV4L=ON -D WITH_GSTREAMER=ON -D WITH_GSTREAMER_0_10=OFF -D WITH_QT=ON -D WITH_OPENGL=ON -D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.7.0/modules ../opencv-4.7.0

  4. Build OpenCV

    First, make sure the Jetson is configured to run at maximum speed.

    sudo nvpmodel -m 3

    jetson_clocks

    Now start the build process. This step takes a very long time (1h+).

    screen -dmS "build_opencv" make -j4

    It is reccomended to use screen here, especially if operating over an ssh connection. This allows the build process to run in a separate shell where it cannot be accidentally stopped if the network is disconnected or the terminal window is closed. To learn more about using screen, see the man page at https://linux.die.net/man/1/screen

  5. Install OpenCV

    sudo make install

  6. To confirm that the OpenCV install has CUDA support, run jtop and click on the INFO tab. You should see

    OpenCV: 4.7.0 with CUDA: YES

Building and installing librealsense

The instructions below are for librealsense 2.53.1 (latest version supported on Ubuntu 18.04 at time of writing)

  1. Download the librealsense source code release for Ubuntu 18.04 from the official git repo

    https://github.com/IntelRealSense/librealsense/releases

    wget https://github.com/IntelRealSense/librealsense/archive/refs/tags/v2.53.1.zip

    unzip v2.53.1.zip

    rm v2.53.1.zip

    Enter the realsense directory

    cd librealsense-2.53.1

  2. Install the necessary build dependencies

    sudo apt install git libssl-dev libusb-1.0-0-dev libudev-dev pkg-config libgtk-3-dev libglfw3-dev libgl1-mesa-dev libglu1-mesa-dev at -y

  3. Build librealsense

    mkdir build && cd build

    cmake -DBUILD_EXAMPLES=true -DFORCE_LIBUVC=true -DBUILD_WITH_CUDA=true ../

    Again, screen is used because the next step may take a while.

    screen -dmS "build_librealsense" make -j4

  4. Install librealsense

    sudo make install

Building and installing realsense-ros

The final step is to build and install the realsense-ros package with support of the newly built librealsense packages. This has been included in the Rover repo.

catkin_make clean

catkin_make

For more detailed build instructions (not necessary for this repo), see here https://github.com/IntelRealSense/realsense-ros/tree/ros1-legacy#step-2-install-intel-realsense-ros-from-sources

Clone this wiki locally