OpenVINO with NCS2 on Raspberry Pi 3 is taking over 5 seconds to process inference (face detection)

[kenenthone]

I am using the following instructions on a Rasbian Buster Lite installation (without X11)

https://docs.openvinotoolkit.org/latest/openvino_docs_install_guides_installing_openvino_raspbian.ht...

I built a project with one of the sample C++ applications and added OpenCV and NGraph libraries and include directories from the OpenVino toolkit installation that were require to build the project

I have the NCS2 plugged into a USB port on the raspberry pi

I was able to build the project, but it was taking over 5 seconds to run inference face detection sample program

I made sure to add the USB rules: "sudo usermod -a -G users "$(whoami)""

How can I check in OpenCV if it is really using Modivius for the CPU inference, and not the CPU?

Are there any pre-built executables that I can use with the raspberry pi camera to test that its using the NCS2?

Thank you,

[David_C_Intel · Accepted solution]

Hi kenenthone,

Thanks for reaching out. On Rpi the CPU plugin is not supported, so any sample you use should be running inference in the MYRIAD plugin. You can check these samples and demos and try to test them. Make sure you are using the latest OpenVINO™ toolkit version with the latest pre-trained models as well, which you can download from here.

Best regards,

David C.

[David_C_Intel]

Hi kenenthone,

As we have not received a response from you, this thread will no longer be monitored. If you have any additional questions, please submit a new post.

Best regards,

David C.