Building OpenCV4.5.5 form source with enabling Inference Engine not working?

I have successfully built OpenCV from the source without any errors, but when I am using Inference Engine to read net, it shows the following error:

OpenCV(4.5.5-dev) D:\opencv\modules\dnn\src\dnn.cpp:3964: error: (-2:Unspecified error) Build OpenCV with Inference Engine to enable loading models from Model Optimizer. in function ‘cv::dnn::dnn4_v20211220::Net::readFromModelOptimizer’

But when I print cv::getBuildInformation(); it prints the following output:

Inference Engine: YES (2021040200 / 2021.4.2) * libs: C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/lib/intel64/Release/inference_engine.lib / C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/lib/intel64/Debug/inference_engined.lib C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/bin/intel64/Release/inference_engine.dll / C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/bin/intel64/Debug/inference_engined.dll C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/bin/intel64/Release/inference_engine.dll C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/bin/intel64/Debug/inference_engined.dll * includes: C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/inference_engine/include nGraph: YES (0.0.0+e2a469a) * libs: C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/lib/ngraph.lib / C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/lib/ngraphd.lib C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/lib/ngraph.dll / C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/lib/ngraphd.dll C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/lib/ngraph.dll C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/lib/ngraphd.dll * includes: C:/Program Files (x86)/Intel/openvino_2021.4.752/deployment_tools/ngraph/include

So the output shows it has the Inference Engine functionality to read net, but why it is not working?

[mshabunin]

I’ve not been able to reproduce your issue with OpenCV@4.5.5 or 4.x branch and OpenVINO 2021.4.2 (model, release).

So I built OpenCV (only dnn module):

$ cd build
$ ..\openvino_752\bin\setupvars.bat
	Python 3.8.12
	[setupvars.bat] OpenVINO environment initialized
$ cmake -G"Visual Studio 16 2019" -A x64 -DWITH_INF_ENGINE=ON  -DBUILD_LIST=dnn ..\opencv
	...
	--     Inference Engine:            YES (2021040200 / 2021.4.2)
	--         * libs:                  C:/work/openvino_752/...
	--         * includes:              C:/work/openvino_752/...
	--     nGraph:                      YES (0.0.0+e2a469a)
	--         * libs:                  C:/work/openvino_752/...
	--         * includes:              C:/work/openvino_752/...
	...
$ cmake --build . --config Release

Then build an application:

$ cd ..\build-sample
$ set OpenCV_DIR=..\build
$ cmake -G"Visual Studio 16 2019" -A x64 ..\example
$ cmake --build . --config Release

Then ran it:

$ set PATH=..\build\bin\Release;%PATH%
$ Release\opencv_example.exe
	>>>>>>>>>>>>>> start
	[E:] [BSL] found 0 ioexpander device
	>>>>>>>>>>>>>> read net done successfully

Application is simple:

cout << ">>>>>>>>>>>>>> start" << endl;
cv::dnn::Net net = cv::dnn::Net::readFromModelOptimizer("person-detection-0203.xml", "person-detection-0203.bin");
cout << ">>>>>>>>>>>>>> read net done successfully" << endl;

[alalek]

Avoid using of CMake GUI - changing build configuration on the fly is tricky and it is not validated.

Consider using of CMake command-line only:

• all required flags are specified on the first run (we have reproducible input as much as possible)
• clear CMake cache if dependencies are changed or upgraded