Community
cancel
Showing results for 
Search instead for 
Did you mean: 
s__jq
Beginner
76 Views

Failed to call OpenVINO function to deal with my own caffe model in OpenCV, while successed with the demo model

    By following the lession article and using the code file named 'openvino_fd_myriad.py' ,I successed to call OpenVINO deal with the demo model in OpenCV. 

    But then, I falied to do that by using another caffe model.

    Here are my steps below:

1.Use the mo_caffe.py to translate the caffe model.     ---  successed

python mo_caffe.py --input_model "posea.caffemodel" --output_dir "ir\FP16" --data_type FP16 --model_name posea --scale_value image[255.0]

2.(python)  call OpenVINO function to deal with my own caffe model in OpenCV.    --- failed, got errors

(1) Code:

inHeight = 384
inWidth = int((inHeight/frameHeight)*frameWidth)
inpBlob = cv2.dnn.blobFromImage(image1, size=(inWidth, inHeight), ddepth=cv2.CV_8U)
net.setInput(inpBlob)
output = net.forward()

(2) Code execution results:


Traceback (most recent call last):
  File "poseN.py", line 198, in <module>
    output = net.forward()
cv2.error: OpenCV(4.0.1-openvino) /home/jenkins/workspace/OpenCV/OpenVINO/build/opencv/modules/dnn/src/op_inf_engine.cpp:553: error: (-215:Assertion failed) Failed to initialize Inference Engine backend: AssertionFailed: inputSize % 2 == 0 in function 'initPlugin'
 

    Can anyone help?  Please ...

 

0 Kudos
3 Replies
s__jq
Beginner
76 Views

( I use the Intel Neural Compute Stick 2 in this case. )

Shubha_R_Intel
Employee
76 Views

Dear jq, 

honestly it's hard to tell what happened in your code just by reading your snippets. But please study carefully the end2end_video_analytics_opencv sample within inference_engine/samples, which seems similar to your example. Likely you made a coding mistake somewhere.

Hope it helps,

Thanks,

Shubha

Shao__Jingqiang
Beginner
76 Views

Dear Shubha:

    Thanks for your reply!   I'll study the sample you recommended. 

    And I found the model "human_pose_estimat..." is incluede in the intel open model zoo, which is

what I want.

    Unfortunately, that model is labeled "Not supportted by  the Intel Neural Compute Stick 2"(NCS2)

as  I understand.

    I hope my understanding is not right, or there is another way to run the model on the NCS2. Could

you tell me about that ?

    Thanks a lot.

    jq

 

Reply