mvNCCompile help

[RDeBo]

Hi,

I have created several custom models. I am trying to deploy them to an NCS 1 and it isn't working.

With the command

mvNCCompile network.prototxt -w network.caffemodel -s MaxNumberOfShaves -in InputNodeName -on OutputNodeName -is InputWidth InputHeight -o OutputGraphFilename

What does 'network.caffemodel' correspond to in a TensorFlow object detection network? All I can see mentioned is .prototxt files; but these are not used in object detection.

On this subject, what is -in also?

I am trying :

mvNCCompile model.ckpt -s 1 -in=frozen_inference_graph.pb -on=node -o graph.graph

but it exists with error 15 -

[Error 15] Toolkit Error: Parser Not Supported: None

Can anyone help please?

[JesusE_Intel · Accepted solution]

Thanks for reaching out!

Try using the following format to compile your network into a Movidius graph file.

mvNCCompile -s <number of shaves> <name of the model>_frozen.pb -in=input -on=output

-in parameter is expecting the name of your input node

-on parameter is expecting the name of your output node

You can find additional information and examples here.

Regards,

Jesus