Models face-detection-0205 and face-detection-0206 showing errors

[campos537]

Hey guys! I'm trying to run the face-detection-0205 and the face-detection-0206 but i wasn't able to get the proper output result, instead of that i got:

terminate called after throwing an instance of 'InferenceEngine::details::InferenceEngineException'
  what():  [PARAMETER_MISMATCH] Failed to set output Blob. Dimensions mismatch.
Aborted (core dumped)

I'm using OpenCV 4.5.1 with OpenVINO 2021.2 with the following code:

cv::dnn::Net detect = cv::dnn::readNet(
      path + "/face-detection-0206/deploy.xml", path + "/face-detection-0206/model.bin", "DLDT");
  detect.setPreferableBackend(2);

  cv::VideoCapture cap(argv[2]);  // cap(0);

  cv::Mat frame;
  int total_frames = 0;
  float total_time = 0;
  while (true) {
    cap >> frame;

    if (frame.empty()) {
      std::cerr << "End of stream" << std::endl;
      exit(0);
    }
    detect.setInput(cv::dnn::blobFromImage(frame, 1.0, cv::Size(640, 640),
                                           cv::Scalar(), false, false, CV_32F),
                    "image");

    cv::Mat a;
    detect.forward("labels");

Also i get the same problem with cv::dnn::DetectionModel

[vladimir-dudnik]

@campos537 we'll look at OpenCV DNN issue, but meantime you may test these models with Open Model Zoo python object_detection_demo.
@l-bat could you please take a look at the issue from OpenCV DNN point of view

[campos537]

@vladimir-dudnik thanks, i will test here with the object_detection_demo and check the results!

[campos537]

@vladimir-dudnik i tested with the latest object_detection_demo using the command line below:
./object_detection_demo -i 0 -m /media/crystal/main_files/Repositorios/eye_is_looking/face_heavy/deploy.xml -at ssd
i got this output here:

[ INFO ] InferenceEngine: 	API version ......... 2.1
	Build ........... 2021.2.0-1877-176bdf51370-releases/2021/2
[ INFO ] Parsing input parameters
[ INFO ] Reading input
[ INFO ] Loading Inference Engine
[ INFO ] Device info: 
[ INFO ] 	CPU
	MKLDNNPlugin version ......... 2.1
	Build ........... 2021.2.0-1877-176bdf51370-releases/2021/2
Loading network files
[ INFO ] Batch size is forced to 1.
[ INFO ] Checking that the inputs are as the demo expects
[ INFO ] Checking that the outputs are as the demo expects
[ ERROR ] [ ERROR ] This demo accepts networks having only one output

[vladimir-dudnik]

@campos537 yeah, but what the model did you use (/media/crystal/main_files/Repositorios/eye_is_looking/face_heavy/deploy.xml)?

Open Model Zoo demo developed and validated on OMZ models, each demo has models.lst file, which you can use as a parameter to model downloader to get models which validated with this demo.

Note, if you still interested to test face-detection-0205/0206 models, they supported in python variant of object_detection_demo.

[campos537]

sorry i changed the name but is the face-detection-0206. Hmm let me test with the python variant to check! Thanks @vladimir-dudnik .

[campos537]

Just an update, i tested with python example and worked normally, any reason it don't work with c++ code?

[vladimir-dudnik]

@campos537 face-detection-0205/0206 have two output layers. We did not modified C++ demo to support this yet.

[campos537]

Thanks for the explanation @vladimir-dudnik! I tested the two output layers "labels" and "boxes" but for some reason didn't worked with C++ OpenCV dnn don't know why.

[vladimir-dudnik]

@campos537 those two a new models, introduced in OpenVINO 2021.2. It might be OpenCV DNN was not updated yet to support them.