Using a Third-party Package (Anaconda) for compiling

1

The openCV tutorial installation notes at OpenCV: OpenCV installation overview indicate that one could use a pre-built binary installation like Anaconda. It works, but it is a bit awkward. You can do:

conda create -n opencv_comp opencv pkg-config
conda activate opencv_comp

and then something like:

g++ -o tutorial_charuco_create_detect `pkg-config --cflags --libs opencv` tutorial_charuco_create_detect.cpp
 LD_LIBRARY_PATH=`pkg-config --variable=libdir opencv` ./tutorial_charuco_create_detect -c=1  # create a charuco pattern in BoardImg.jpg

Anaconda environments allow you to set a LD_LIBRARY_PATH variable for the environment:

conda env config vars set LD_LIBRARY_PATH=`pkg-config --variable=libdir opencv

# then...
conda deactivate
conda activate opencv_comp

 # get hello_world_opencv.cpp from gist

g++ -o hello_world `pkg-config --cflags --libs opencv` hello_world_opencv.cpp

./hello_world_opencv

The pkg-config features enable folks to use a binary distribution for opencv experimentation without installing a full source installation. I imagine other third-party binary installs besides Anaconda have similar features.

I opened an issue at Compiling tutorial_charuco_create_detect.cpp using opencv_contrib in Anaconda · Issue #2948 · opencv/opencv_contrib · GitHub since the Third-party binary installs seemed at least suggested in the docs

Is there a place to raise an issue with the tutorial doc specifically? It might be nice to add a couple lines under the Third-party Packages ‘Conda’ bullet point with hints on how to use it:

# Anaconda Example:
conda create -n opencv_comp opencv pkg-config
conda activate opencv_comp
 # get hello_world_opencv.cpp from gist

g++ -o hello_world `pkg-config --cflags --libs opencv` hello_world_opencv.cp

LD_LIBRARY_PATH=`pkg-config --variable=libdir opencv` ./hello_world_opencv
2

Actually pkg-config method is deprecated, it is recommended to use cmake. It requires to create separate directory and a project file, but is more robust and better supported method of using OpenCV. I doubt this instruction can be added to the documentation.
Check this page for recommended usage method: OpenCV: Using OpenCV with gcc and CMake (though it is quite old and can be improved).

Also, there is more recent version of OpenCV available in the conda-forge channel: https://anaconda.org/conda-forge/opencv

Perhaps separate tutorial covering OpenCV Conda package usage on various platform can be added to the documentation. It would be useful.