Failed to initialize Inference Engine backend , opencv dnn

[rohit12]

I am trying to use opencv dnn module to perform object detection. I am using movidius neural compute stick with openvino toolkit. 

Code is as below:

# load the DNN model
model = cv2.dnn.readNet(model=model_path, config=config_path, framework='TensorFlow')

model.setPreferableBackend(cv2.dnn.DNN_BACKEND_INFERENCE_ENGINE)

model.setPreferableTarget(cv2.dnn.DNN_TARGET_MYRIAD)

Getting an error message: 

output = model.forward()
cv2.error: OpenCV(4.5.3-openvino) C:\jenkins\workspace\windows6296ace0\opencv\modules\dnn\src\ie_ngraph.cpp:738: error: (-2:Unspecified error) Failed to initialize Inference Engine backend (device = MYRIAD): Failed to allocate graph: NC_ERROR in function 'cv::dnn::InfEngineNgraphNet::initPlugin'

Could someone please help me resolve it. Thanks in advance.

[Vladimir_Dudnik]

@rohit12 according to error message, runtime was not able to allocate network graph on device. Potentially your model might be too big to be loaded into NCS2 device.

[rohit12]

Hello @Vladimir_Dudnik , 

is it necessary that the model should always be in intermediate representation (IR) format that is .xml and .bin to execute it on NCS2?

In the above query I am passing .pb and .pbtxt files which is tensorflow format

[Peh_Intel]

Hi rohit12,

Thank you for your question. If you need any additional information from Intel, please submit a new question as this thread is no longer being monitored.

Regards,

Peh

[Vladimir_Dudnik · Accepted solution]

Hello @rohit12, that's rights, in case of OpenVINO backend in OpenCV DNN you have to supply model in IR form only

Regards,
  Vladimir

[KSehairi]

Hello
you have to copy .dll files located on …Intel\openvino_2022.3.1\runtime\bin\intel64\Release to your cv2 folder or add this folder to system path