DepthAI C++ build with OpenCV support fails

MellevanderHeijden

I am experiencing issues building DepthAI core C++ library with OpenCV support enabled on my Ubuntu 22.04 system with CUDA-enabled OpenCV 4.13.0 installed from source.

Despite successful detection of OpenCV in cmake logs and correct linking of OpenCV libs in the DepthAI shared library, compiling any DepthAI calls that require OpenCV support - ie. ImgFrame::getCvFrame - result in static assertion errors indicating "Library not configured with OpenCV support".

Environment:

Ubuntu 22.04.4 LTS
GCC 12.3.0
NVIDIA Driver version : 580.82.07
NVIDIA CUDA: 13.0
OpenCV 4.13.0-dev
- built from source with CUDA support, installed under /usr/local
DepthAI Core: [commit hash, tag version]
- v3.0.0-rc.4
- hash: 4adc1dfe3fccb11184ccd05a673673e74e0f30d9
cmake 3.22.1

Steps to Reproduce:

Clean install OpenCV from source with CUDA, installed to /usr/local
Clone DepthAI core, submodules update
Configure DepthAI with cmake: -D OpenCV_DIR=/usr/local/lib/cmake/opencv4 -D DEPTHAI_OPENCV_SUPPORT=ON
Build and install DepthAI core
Attempt to build OpenFrameworks project invoking DepthAI OpenCV APIs

Observed Behavior:

CMake logs show DEPTHAI_OPENCV_SUPPORT enabled
ldd on libdepthai-core.so confirms linkage with OpenCV libraries
Compiler errors persist when calling OpenCV-dependent APIs, with static assertion: Library not configured with OpenCV support

Logs:

I prepared all the logs to attach to this ticket, but I'm not allowed to upload any. "Uploading files of this type is not allowed." I've tried numerous extensions and they all fail. If you need access to certain log files, please ask and provide a method of transfer. You can find the cmake config flags I've used to build opencv and depthai below.

opencv cmake -D flags:

cmake -D'CMAKE_BUILD_TYPE=Release' \

-D'CMAKE_INSTALL_PREFIX=/usr/local' \

-D'OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules' \

-D'WITH_CUDA=ON' \

-D'WITH_CUDNN=ON' \

-D'OPENCV_DNN_CUDA=ON' \

-D'CUDA_ARCH_BIN=8.9' \

-D'CUDA_HOST_COMPILER=/usr/bin/g++' \

-D'OPENCV_GENERATE_PKGCONFIG=ON' \

-D'CPU_BASELINE=SSE3' \

-D'CMAKE_VERBOSE_MAKEFILE=ON' \

-D'BUILD_opencv_python3=ON' \

-D'CURL_INCLUDE_DIR=ls -l /usr/include/x86_64-linux-gnu/curl' \

-D'CURL_LIBRARY=/usr/lib/x86_64-linux-gnu/libcurl.so' \

-D'BUILD_EXAMPLES=ON' ..

depthai cmake -D flags:

cmake -B build -S . \

-D'BUILD_SHARED_LIBS=ON' \

-D'CMAKE_INSTALL_PREFIX=/usr/local/' \c

-D'OpenCV_DIR=/usr/local/lib/cmake/opencv4' \

-D'CMAKE_VERBOSE_MAKEFILE=ON' \

-D'DEPTHAI_OPENCV_SUPPORT=ON' ..

Please advise on resolving this issue or inquire about additional information if needed.

Thank you in advance.

jakaskerl

MellevanderHeijden
IIRC there is another .so that is generated called depthai-opencv or something similar. It should have opencv support.

Thanks,
Jaka

MellevanderHeijden

jakaskerl Thanks for your reply.

libdepthai-core.so is the only depthai related .so present in my /usr/local/lib/ folder.
Using 'find' yields nothing for 'depthai-opencv.so'

MatevzMorato

Hey @MellevanderHeijden, sorry for the delay.

The reason why the static asserts are happening is likely due to the way you're integrating the library, I assume it's not with CMake?

We don't support non CMake integration nicely yet, you will have to provide the defines (DEPTHAI_OPENCV_SUPPORT) which match how you compiled the library manually for now.

MatevzMorato

If possible, we kindly recommend doing the integration with CMake as you can see here:
luxonis/depthai-core-example

MellevanderHeijden

Hey MatevzMorato ,

Thanks for your reply. You are correct, I am not using cmake to compile my project. I'm using OpenFrameworks, which comes with its own build tools. I understand it is possible to build the OpenFrameworks project using cmake instead, although not straightforward. From your message, I understand properly integrating depthai using OpenFrameworks' build tools is not supported at the moment. I will start working on the former option and follow up here if I run into any problems.

Thanks for your assistance.