OpenVINO pre trained models don't work in Raspberry pi3 with NCS

[Alba]

Dear all,

I am trying to use the pre-trained model "person-detection-retail-0013" with the Raspberry pi3 and NCS. I have installed OpenVINO 2020.3.

When I call the "load_network" function to create an executable network, it always fails in an error like this:

RuntimeError: Check 'input_shape_product.get_length() == shape_size(m_output_shape)' failed at /home/jenkins/agent/workspace/private-ci/ie/build-linux-debian_9_arm/b/repos/closed-dldt/ngraph/src/ngraph/op/reshape.cpp:100:
While validating node 'v0::Reshape Reshape_17724(PriorBoxClustered_15181[0]:f32{2,4042567040}) -> (dynamic?)':
Product of output shape dimensions does not match product of argument shape dimensions (output shape: Shape{1, 2, 4042567040}, argument shape: {2,4042567040}).

Sometimes it also outputs "aborted" instead.

I try also with different pre-trained models but the error persist.

How can I fix it?

Thank you!

[AndrewG_Intel]

Hello Alba

Thank you for posting on the Intel® communities.

We noticed that your inquiries are regarding "Intel® Distribution of OpenVINO™ toolkit". We have a forum for these specific products and issues. We are moving this thread to the ‘Intel® Distribution of OpenVINO™ Toolkit forum’, so you can get answered more quickly.

Best regards,

Andrew G.

Intel Customer Support Technician

[David_C_Intel · Accepted solution]

Hi Alba,

Thanks for reaching out. It might be due to an incompatibility with the latest IR versions and OpenVINO™ toolkit for Raspbian. Please try to use the 2019 R3 version models. If the error persists, please share the script you are using for us to test from our end.

Regards,

David C.

[Alba]

Hello,

With the 2019 R3 models works.

Thank you!

[David_C_Intel]

Hi Alba,

Intel will no longer monitor this thread since this issue has been resolved. If you need any additional information from Intel, please submit a new question. 

Regards,

David C.