Error: Yolov5 Openvino model run on Raspberry Pi 4 with NCS2

[Bendan]

Hi all,

I wish to run my best_openvino_model on my Raspberry Pi 4 with NCS2. I exported the openvino model using the official way, and facing error: "requirements: YOLOv5 requirement "openvino" not found" and "No module named "openvino"" as captured in the figure below.

After installing openvino==2021.4.582, I tried the face-detection-adas-0001.xml, and it ran smoothly without error. The version of best.xml generated is in IR 11 format.

Anyone has any idea on this error? Thank you. 

 

[Wan_Intel]

Hi Bendan,

Thanks for reaching out to us.

OpenVINO™ toolkit version 2021.4.582 does not support IR 11 format. Please use IR 10 format when using OpenVINO™ toolkit version 2021.4.582.

Regards,

Wan

[Bendan]

Hi Wan,

Thank you for replying me. However, I tried different ways to convert to IR 10 format, but the output is still in IR 11 format. Can you provide some documentations/resources that convert a Pytorch model into IR 10 format on OpenVINO™ toolkit version 2021.4.582? 

Thank you.

[Wan_Intel]

Hi Bendan,

You may convert your model using OpenVINO™ toolkit version 2021.4. Please refer to Converting a PyTorch Model for more information.

Regards,

Wan

[Bendan]

Hi Wan,

Thank you for the information!

Best regards,

Leong Rong Chuan

[Wan_Intel]

Hi Bendan,

If you need additional information from Intel, please submit a new question as this thread will no longer be monitored.

Best,

Wan

[BarryFu]

Hi, Bendan

           Could you help me to list the install information such as Raspbian version, the web document which was provided step-by-step method and related packages on Raspberry pi 4? I used  2019-09-26-raspbian-buster, l_openvino_toolkit_runtime_raspbian_p_2021.4.582.tgz and followed the website of https://docs.openvino.ai/2021.4/openvino_docs_install_guides_installing_openvino_raspbian.html 

on my Raspberry pi 4 and NCS2. I run openVINO inference with face-detection-adas-0001.xml model unsuccessful.