Since I ran into some really painfull setup issues I've written down some stuff I learned along the way. This repository contains some tips to setup your development environment for the exercises in the Augemented Reality lecutre. Recommendations about building from source or using a docker image (recommended to save time).

If your distribution of choice does not have opencv in its repositories you have to install it using homebrew or build it from source.

Relevant Links

• https://docs.opencv.org/4.x/d7/d9f/tutorial_linux_install.html
• https://docs.opencv.org/4.x/d0/da7/videoio_overview.html

Notes on building from source

• You need both the opencv and opencv_contrib repos
• Make sure ffmpeg is installed
• Check output of the cmake configuration stage to make sure everything you need is included
  • VideoIO: FFMPEG On
  • Modules: core,imgcodecs,imgproc,videoio,objdetect,video,tracking
• If the build refuses to include ffmpeg, try removing it, installing pkg-config and then reinstalling ffmpeg (potentially alongside libavformat-dev libavcodec-dev libswscale-dev)
• Using make install the .so files are only installed to /usr/local/lib which is not on the default linker path, therefore you either need to add it to your linker path (LD_LIBRARY_PATH for gcc) or copy all required .so files to your projects bin folder
• You might need to update your c++ header include path (CPLUS_INCLUDE_PATH)
• Build only selected modules using: -DBUILD_LIST=<module-list>
• Enable ffmpeg video backend: -DWITH_FFMPEG=ON
• Enable dynamically linked ffmpeg: -DWITH_FFMPEG=ON -DVIDEOIO_PLUGIN_LIST=ffmpeg
• Inside your build directory

cmake -DOPENCV_EXTRA_MODULES_PATH=path/to/opencv_contrib-\<version>/modules -DBUILD_LIST=\<module-list> -DWITH_FFMPEG=ON \<-DVIDEOIO_PLUGIN_LIST=ffmpeg> path/to/opencv-\<verison>

• Exmaple Usage

cmake -DOPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.9.0/modules -DBUILD_LIST=core,imgcodecs,imgproc,videoio,objdetect,video,tracking -DWITH_FFMPEG=ON -DVIDEOIO_PLUGIN_LIST=ffmpeg ../opencv-4.9.0

Download the docker image

docker pull kauzoo/ar-docker:latest  

OpenCV is built from source, using only modules required for lecture and dependencies. Ffmpeg is used as dynamicaly linked VideoIO backend (libopencv_videoio_ffmpeg.so). Versions and module list can be tweaked using docker build-arguments (See build).

• OpenCV-version: 4.9.0 (default)
• OpenCV_contrib-version: 4.9.0 (default)
• Godot-version: 4.4.0 (default)
• Godot-cpp-version: 4.4 (default)

You can use the docker container as your development environment for compilation and code editing. The godot editor should be run on your host machine (running it inside the container doesn't work properly). The .devcontainer.json file can be copied into the root of your project (next to sourcecode src/ and godotproject (probably demo) directory) to set your project up as a vscode dev container. Open your project folder with vscode and you should be prompted to Reopen in Container.

1. Pull docker image

docker pull kauzoo/ar-docker:latest

2. In .devcontainer.json adjust

   • WORKSPACE_FOLDER = project directory
   • GODOT_PROJECT_NAME = godot project directory (contains project.godot)

3. Copy .devcontainer.json to the root folder of your project (next to the SConstruct file etc.)

4. Open the project folder in vscode

5. Run >Dev Containers: Reopen in Container (Ctrl+Shift+P)

6. (First Usage) Copy godot-cpp to your project

cp -r /home/ardocker/export/godot-cpp "/workspaces/$WORKSPACE_FOLDER/godot-cpp"

6. (First Usage) Copy libs to your project

cp -r /usr/lib/libopencv_* "/workspaces/"$WORKSPACE_FOLDER/$GODOT_PROJECT_NAME/bin/

7. Build from inside the container

scons platform=linux

8. On Host machine: Open godot project like usual

If you just want to export all relevant files you can use the export.sh script. The latest version of the image will be pulled with default settings and the files will be exported. (Hint: Docker must be installed).

• opencv-libs (*.so): -> ./ar-docker-export/opencv/lib
• opencv-headers: -> ./ar-docker-export/opencv/include
• godot-cpp: -> ./ar-docker-export/godo-cpp

chmod +x ./export.sh && ./export.sh

• opencv_version=4.9.0
• godot_version=4.4-stable
• gd_cpp_version=4.4
• opencv_build_list=core,imgcodecs,imgproc,videoio,objdetect,video,tracking

Example Usage

docker build --build-arg opencv_version=4.9.0 --build-arg godot_version=4.4-stable --build-arg gd_cpp_version=4.4 --build-arg opencv_build_list=core,imgcodecs,imgproc,videoio,objdetect,video,tracking .

• If godot produces the Couldn't open camera. error ffmpeg is proabably not enabled as a backend