Compile OpenCV for Linux devices

OpenCV is a library of programming functions mainly aimed at real-time computer vision.

Here I introduced some ways to compile OpenCV with Contrib for different devices like Jetson TX2, Raspberry Pi 4B and x86_64 PC.

Install dependencies

  • Update source and upgrade packages, clean and remove unused packages
1
2
3
4
sudo apt update
sudo apt upgrade -y
sudo apt clean
sudo apt autoremove -y
  • Install python3 , pip3 and necessary wheels for Python support
1
2
sudo apt install python3-dev python3-pip -y
sudo pip3 install cython setuptools numpy
  • Install third-party software libraries that require to build
1
2
3
4
5
6
7
8
sudo apt install gcc-arm* build-essential cmake git unzip pkg-config -y
sudo apt install protobuf-compiler libgflags-dev libgoogle-glog-dev -y
sudo apt install libjpeg-dev libpng-dev libtiff-dev -y
sudo apt install libavcodec-dev libavformat-dev libavutil-dev libxvidcore-dev libswscale-dev libx264-dev -y
sudo apt install liblapack-dev libxine2-dev gfortran -y
sudo apt install libpostproc-dev libswscale-dev -y
sudo apt install libmpikmeans-dev libmpikmeans1 mpi-default-dev -y
sudo apt install libgtk-3-dev -y

For a special package called libjasper-dev, which only available before Ubuntu Xenial, we add Xenial sources(Here for Ubuntu restricted systems, we could use Tsinghua mirrors)

1
2
3
sudo echo "deb http://mirrors.tuna.tsinghua.edu.cn/ubuntu-ports/ xenial main multiverse restricted universe" >> /etc/apt/sources.list
sudo apt update
sudo apt install libjasper-dev -y

Custom configuration

For different hardwares, it’s better to compile acceleration enabled OpenCV.

GPU with CUDA

Like Jetson TX2, which hardware has standalone GPU support, that we could enable GPU acceleration by add its CUDA libs while compiling.

  • Install CUDA from dpkg
1
2
3
sudo dpkg -i <CUDA deb path>
sudo dpkg -i <CuDnn deb path>
sudo dpkg -i <NightSystems deb path>
  • Confirm installation using apt
1
2
3
4
5
6
7
8
apt search cuda
sudo apt install <CUDA name>

apt search cudnn
sudo apt install <CuDnn name>

apt search NightSystems
sudo apt install <NightSystems>

Low-memory devices

For the devices which has small RAM size like Raspberry Pi 4B (2G), it’s recommended that to use large swap.

  • Create 2G or larger blank file to disk
1
sudo fallocate -l 2G /swapfile
  • Protect memory, only the root user should be able to write and read the swap file
1
sudo chmod 600 /swapfile
  • Use the mkswap utility to set up the file as Linux swap area
1
sudo mkswap /swapfile
  • Enable the swap and verify
1
2
sudo swapon /swapfile
sudo swapon --show

To make the change permanent open the /etc/fstab file and append the following line by this command

1
sudo echo "/swapfile swap swap defaults 0 0" >> /etc/fstab

Furthermore, you could custom the swappiness value. Swappiness is a Linux kernel property that defines how often the system will use the swap space. Swappiness can have a value between 0 and 100. A low value will make the kernel to try to avoid swapping whenever possible, while a higher value will make the kernel to use the swap space more aggressively.

  • Set your swappiness, the default is 60
1
sudo sysctl vm.swappiness=60

To remove swap file, deactivate the swap and delete swapfile

1
2
3
sudo swapoff -v /swapfile
sudo sed -i ‘/swapfile/d’ /etc/fstab
sudo rm /swapfile

Gstreamer

GStreamer is a library for constructing graphs of media-handling components. The applications it supports range from simple Ogg/Vorbis playback, audio/video streaming to complex audio (mixing) and video (non-linear editing) processing.

1
sudo apt install libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev -y

Threading building blocks

Threading Building Blocks (Intel® TBB) makes parallel performance and scalability accessible to software developers who are writing loop- and task-based applications. Build robust applications that abstract platform details and threading mechanisms while achieving performance that scales with increasing core count.

1
sudo apt install libtbb2 libtbb-dev -y

Handing medias

The v4l-utils are a series of packages for handling media devices, libdc1394 is a library that provides a complete high level application programming interface (API) for developers who wish to control IEEE 1394 based cameras that conform to the 1394-based Digital Camera Specifications (also known as the IIDC or DCAM Specifications).

1
sudo apt install v4l-utils libv4l-dev libdc1394-22-dev -y

BLAS

BLAS (Basic Linear Algebra Subroutines) is a set of efficient routines for most of the basic vector and matrix operations. They are widely used as the basis for other high quality linear algebra software, for example lapack and linpack.

1
sudo apt install libopenblas-dev libatlas-base-dev libblas-dev -y

QT5

Qt is a cross-platform C++ application framework. Qt’s primary feature is its rich set of widgets that provide standard GUI functionality.

1
sudo apt install qt5-default -y

HDF5

Hierarchical Data Format (HDF) is a set of file formats (HDF4, HDF5) designed to store and organize large amounts of data.

1
sudo apt install libhdf5-serial-dev hdf5-tools libhdf5-dev -y

Tesseract

Tesseract is an open source Optical Character Recognition (OCR) Engine. It can be used directly, or (for programmers) using an API to extract printed text from images. It supports a wide variety of languages.

1
sudo apt install libtesseract-dev tesseract-ocr -y

Eigen

Eigen is a versatile, fast, reliable and elegant C++ template library for linear algebra: matrices, vectors, numerical solvers, and related algorithms.

1
sudo apt install libeigen3-dev -y

Download sources

OpenCV Release - GitHub

OpenCV Contrib Release - GitHub

  • Source code can be founded on GitHub, then we download it.
1
2
3
4
pushd /tmp

wget -O opencv.zip https://github.com/opencv/opencv/archive/4.5.0.zip
wget -O opencv_contrib.zip https://github.com/opencv/opencv_contrib/archive/4.5.0.zip
  • Unzip source and enter OpenCV source, create build folder.
1
2
3
4
unzip opencv.zip opencv_contrib.zip
pushd opencv-*.*.*
mkdir build
pushd build

Optimized compiling

CMake is a cross-platform build system generator. Projects specify their build process with platform-independent CMake listfiles included in each directory of a source tree with the name CMakeLists.txt. Users build a project by using CMake to generate a build system for a native tool on their platform.

Here I compile OpenCV with OpenCV Contrib by default:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
cmake -D CMAKE_BUILD_TYPE=RELEASE \
      -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib-*.*.*/modules \
      -D WITH_OPENMP=ON \
      -D WITH_TBB=ON \
      -D BUILD_TBB=ON \
      -D BUILD_TIFF=ON \
      -D WITH_GSTREAMER=ON \
      -D WITH_V4L=ON \
      -D WITH_LIBV4L=ON \
      -D OPENCV_ENABLE_NONFREE=ON \
      -D INSTALL_C_EXAMPLES=OFF \
      -D INSTALL_PYTHON_EXAMPLES=OFF \
      -D BUILD_NEW_PYTHON_SUPPORT=ON \
      -D BUILD_opencv_python3=TRUE \
      -D WITH_QT=ON \
      -D BUILD_EXAMPLES=OFF \
      ..

Arch

While the devices has both AArch (ARMv8) CPU and software environment (64bit kernel and 64bit compiler), such as Jetson TX2, Raspberry Pi 4B with 64bit OS, there would be something different for CMake flags.

  • AArch32
1
2
-D ENABLE_NEON=ON \
-D ENABLE_VFPV3=ON \
  • AArch64
1
-D ENABLE_NEON=ON \

GPU

We have noticed that in most scenarios OpenCV utilizes CPU, which doesn’t always guarantee our the desired performance.

  • CUDA
1
2
3
4
5
-D WITH_CUDA=ON \
-D WITH_CUBLAS=ON \
-D CUDA_FAST_MATH=ON \
-D ENABLE_FAST_MATH=ON \
-D WITH_OPENGL=ON \

Java API

Build OpenCV Java API requires Java runtime, make sure jdk is installed first, and then reconfigure CMake flags.

1
sudo apt install openjdk-11-jdk ant -y

Note: Compile OpenCV with Contrib may require unblocked Internet connection.

Compile and Test

Normally we build from source code by sudo make, but it seems always slow in a more awkward way. Pay attention to -j [N], --jobs[=N], note that sudo make -j 4 means build using 4 threads as 4 cores, then we got this command which makes the build process much faster.

1
make -j "$(grep -c ^processor /proc/cpuinfo)"

If there’s no errors displayed, it means you compiled OpenCV successfully. Then, we install and linking.

1
2
sudo make install
sudo ldconfig

For Jetson TX2, Here’s an example to test onboard camera and if Gstreamer works.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
#include <opencv2/core.hpp>
#include <opencv2/videoio.hpp>
#include <opencv2/highgui.hpp>

int main() {
    cv::VideoCapture cap;
    cv::Mat          frame;

    cap.open("nvarguscamerasrc ! video/x-raw(memory:NVMM), width=(int)720, height=(int)480, format=(string)NV12, framerate=(fraction)30/1 ! nvvidconv ! video/x-raw, format=BGRx ! videoconvert ! video/x-raw, format=BGR ! appsink");
    while (cap.isOpened()) {
        cap.read(frame);
        if (frame.empty()) { continue; }
        cv::imshow("frame", frame);
        if (cv::waitKey(5)) { break; }
    }

    return 0;
}