Reshape issue while running net.forward()

[ArushikaBansal]

While running notebook 'multi-person-openpose.ipynb' encountered an issue in the code

t = time.time()
net = cv2.dnn.readNetFromCaffe(protoFile, weightfile)
net.setPreferableBackend(cv2.dnn.DNN_BACKEND_OPENCV)
net.setPreferableTarget(cv2.dnn.DNN_TARGET_CPU)

if device == "cpu":
net.setPreferableBackend(cv2.dnn.DNN_TARGET_CPU)
print("Using CPU device")
elif device == "gpu":
net.setPreferableBackend(cv2.dnn.DNN_BACKEND_CUDA)
net.setPreferableTarget(cv2.dnn.DNN_TARGET_CUDA)
print("Using GPU device")

Fix the input Height and get the width according to the Aspect Ratio
inHeight = 368
inWidth = int((inHeight/frameHeight)*frameWidth)

inpBlob = cv2.dnn.blobFromImage(image1, 1.0 / 255, (inWidth, inHeight),
(0, 0, 0), swapRB=False, crop=False)

net.setInput(inpBlob)
output = net.forward()
print("Time Taken = {}".format(time.time() - t))

---

error Traceback (most recent call last)
in <cell line: 23>()
21 # inpBlob = np.squeeze(inpBlob,0)
22 net.setInput(inpBlob)
---> 23 output = net.forward()
24 # print("Time Taken = {}".format(time.time() - t))

error: OpenCV(4.7.0) /io/opencv/modules/core/src/matrix.cpp:1240: error: (-209:Sizes of input arguments do not match) Requested and source matrices have different count of elements in function 'reshape'

Please help me to solve this issue. Thanks in advance :)

[brmarkus]

Looks like the code doesn't match to the model (anylonger).
Which version of the code/notebook/repo and which version of the model do you use? Do you use the original model of the sample, or used your own retrieved one?
Can you add a few print-log-messages of the provided input-shape and of the expected input-shape? You might check the model via a model-viewer like "Netron" and check the expected input-shape.

[ArushikaBansal]

I have simply run the notebook below:
https://github.com/spmallick/learnopencv/blob/master/OpenPose-Multi-Person/multi-person-openpose.ipynb
with protoFile and weightFile as given. I kept the input also the same.
The Openpose vesion is 4.7.0

[brmarkus]

Now I tried it under MS-Win10 in a Python virt-env using thhe Jupyter notebook as well.
I got a cv2.error: Unknown C++ exception with the instruction net.setInput(inpBlob).

Via Google search I found e.g. this pointer: "https://stackoverflow.com/questions/75334101/layeroutputs-net-forwardoutput-layers-names-cv2-error-unknown-c-exception"

Yes, I had manually installed pip3 install opencv-python, and now manually installed pip3 install opencv-python==4.6.0.66.

With this version I got the Jupyter notebook to run successfully: