1

I searched a while and could not find any documents about deploying the OpenVINO IR model directly to Intel's MyriadX VPU.

My understanding is they both come from Intel, there should be a way to do this.

What can I try in order to solve this?

halfer
  • 19,824
  • 17
  • 99
  • 186
Franva
  • 6,565
  • 23
  • 79
  • 144

2 Answers2

0

To infer the IR with MYRIAD, You may use the OpenVINO official Samples for that purpose. Use the -d MYRIAD parameter to run it on Myriad.

Don't forget to install the Plugin.

This seems to be the same question as this Github thread.

Rommel_Intel
  • 1,369
  • 1
  • 4
  • 8
  • Hi there, thanks for sharing the solution. But I found that there is no code sample for using `.blob` model directly on VPU in Python. Could you please provide the code in the document? – Franva Aug 24 '21 at 09:11
0

If you read the documentation that I had shared previously, you will notice the -h parameter would display everything that could be done with that compile_tool. You will get the answer there actually.

Run the compile_tool: compile_tool.exe -m "model_location\efficientdet-d0-tf\FP16\efficientdet-d0-tf.xml" -d MYRIAD -o "location_you_want_to_store_the_blob\blob_efficientdet\blob_filename_must_have.blob"

Datasheet/Documentation is your best friend now, make sure to read and try to understand them.

Another thing to take note is, anything related with NCS2/MYRIAD/VPU must be in FP16 format.

enter image description here

Rommel_Intel
  • 1,369
  • 1
  • 4
  • 8
  • Hi there, thanks for sharing this, but I'm not asking for how to compile IR model into blob format. Yep, I did run that `-h` and found there are FP32, F16, half and one more format. Btw, the `compile_tool` has a bug there and I have reported it. it should be fixed later. – Franva Aug 25 '21 at 11:08