Error Messages during OpenVINO Inferencing

[DarkHorse]

Hi,

I got this error after I export my trained IR files trained on Intel GETI to a object detection project in Intel OpenVINO.

Below are the error message:

RuntimeError: Check 'std::get<0>(valid)' failed at src/inference/src/ie_core.cpp:1689:
InferenceEngine::Core::LoadNetwork doesn't support inputs having dynamic shapes. Use ov::Core::compile_model API instead. Dynamic inputs are :{ input:'image,image', shape=[?,3,736,992]}

I wonder why the generated IR files cannot run out of the box if I port it over to an object detection project in OpenVINO.

Thanks.

[Zulkifli_Intel]

Hi DarkHorse,

Thank you for reaching out to us.

Are you creating a custom code or you are using an Object Detection Demo from Intel Open Model Zoo? Also, which OpenVINO version did you use?

Based on the error, the model contains a dynamic shape. You can refer to class ov::CompiledModel and Dynamic Shape for more information.

Sincerely,

Zulkifli

[DarkHorse]

Hi Zul,

yes I am using Object Detection Demo from Open Model Zoo. I am using OpenVino 2022. If the IR model generated from Intel GETI run together with the inference project which was generated, it is running fine.

But when we port the IR model from Intel GETI to run in the Object Detection Demo from Open Model Zoo the error starts to show up. I was wondering, why Intel GETI doesn't remove the dynamic layers of the model to make it compatible to run with any project in Open Model Zoo?

Thanks.

[Zulkifli_Intel]

Hi DarkHorse,

This error could be due to the model being converted using an old OpenVINO version. Can you help to check in Geti which versions of Geti and OpenVINO you used?

Sincerely,

Zulkifli

[DarkHorse]

Hi Zul,

Openvino 2023

Intel® Geti™ - 1.5.1

Thanks

[Zulkifli_Intel]

Hi DarkHorse,

Thank you for your patience.

Based on requirements.txt, Intel Geti should be OV 2022.3 and Geti exported models are not applicable to run on the demo scripts on Open Model Zoo. This is because of the OpenVINO training extension (OTX) is used by the Geti model training where you will need to have the wrapper of OTX runtime to run Geti's model inferencing. OTX runtime is not provided in the demo code of Open Model Zoo. Hope this information helps.

Sincerely,

Zulkifli

[Zulkifli_Intel]

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.