How to Set Up OpenCV on Raspberry Pi 4

Expanding the File System

We are expanding the file system for a smoother process of installing OpenCV libraries. So to do this, we need to get to the Raspberry Pi terminal and type the command below.

sudo raspi-config

After typing this command, a configuration menu will open up, select the option Advanced Options and then select Expand Filesystem. After selecting the option, it will ask for confirmation to reboot the system, then select Y or Yes.

It is important to reboot the system after this step so that Raspberry Pi will expand the filesystem then and there.

Updating the System

We need to update the system if any latest packages are available; this also helps for a faster and better installation process. For this, open the terminal and write the below code.

sudo apt-get update && sudo apt-get upgrade

This step will take a few minutes to get the Raspberry Pi to the latest version.

Installing Snapd and CMake.

We need CMake for compiling OpenCV libraries, but for that, we first need to install Snapd, which contains the CMake package. To do that, write the code given below in the terminal.

sudo apt install snapd

Now install CMake using Snapd, and enter the below command.

sudo snap install cmake --classic

This will install CMake. If any error appears while installing CMake, you can increase the GPU Memory.

For this, go to Raspberry Pi Icon, select Preferences, then click on Raspberry Pi Configuration, and hit Enter. Once the dialog box appears, go to Performance and increase the GPU Memory to 256 (if it is less than 256).

Installing Python 3

To install Python 3 on your Raspberry Pi system, enter the below command in your terminal.

sudo apt-get install python3-dev

Downloading and Installing OpenCV Packages

For this, it is recommended not to use the latest version because it won’t be stable; we will require a more stable version. So we will download the OpenCV package and configure it on Raspberry Pi using CMake.

We need to download the OpenCV zip file from the GitHub link using the code below.

wget -O opencv.zip https://github.com/opencv/opencv/archive/4.5.0.zip

Now we need to download the opencv-contrib zip file. The difference between OpenCV and opencv-contrib is that OpenCV has only main module packages while opencv-contrib contains a full package containing the main modules and contrib/extra modules.

We download both zip files so that OpenCV works correctly.

wget -O opencv_contrib.zip https://github.com/opencv/opencv_contrib/archive/4.5.0.zip

And so after downloading these two zip files, we unzip them separately using the command lines shown below.

unzip opencv.zip

unzip opencv_contrib.zip

Now both these files are ready to be used.

Installing NumPy

OpenCV requires support from NumPy to work, so we install NumPy.

pip install numpy

Building a directory

When we unzipped the two zip files, two new directories were created. We need to add a new directory called Build in the opencv-4.5.0 directory.

cd ~/opencv-4.5.0
mkdir build
cd build

Compiling CMake and OpenCV

Now we compile CMake for OpenCV, but first, we need to check if we are in the path ~/opencv-4.5.0/build. If the path is right, we can copy and paste the commands given below.

cmake -D CMAKE_BUILD_TYPE=RELEASE \
-D CMAKE_INSTALL_PREFIX=/usr/local \
-D OPENCV_EXTRA_MODULES_PATH=~/opencv_contrib-4.0.0/modules \
-D ENABLE_NEON=ON \
-D ENABLE_VFPV3=ON \
-D BUILD_TESTS=OFF \
-D WITH_TBB=OFF \
-D INSTALL_PYTHON_EXAMPLES=OFF \
-D BUILD_EXAMPLES=OFF ..

In the end, it shows Configuration Done and Generating Done, which means it got configured correctly without any errors.

For compiling OpenCV, we use the code below in the same path.

make -j4

This will compile OpenCV completely, but it will take time. However, if this doesn’t compile, then we can use make -j1; this will take more time to compile than j4, but it will be stable.

Installing OpenCV libraries

Install the OpenCV libraries using the following code snippet.

sudo apt-get install libopencv-devpython-opencv

OpenCV Testing on Raspberry Pi

Finally, in the terminal on the same build path mentioned above, type python and then import cv2, as shown below.

python
import cv2
cv2.__version__

If the cv2 is imported and the version appears as expected, the installation and setup are done successfully.