what's the difference between nGraph & OpenVINO optimizer?

[chang__john]

As title:

Are nGraph & openVINO optimizer both optimizers for a given trained graph?

If NOT so, what is the difference between them?

I checked the readme of nGraph and it claims the support for Intel Movidius™ Myriad is "coming soon".

But I think to maintain two optimizers makes no sense at all.

So I wonder what's the optimizer behind OpenVINO? is it the same as nGraph?

Which one is better when using Intel Movidius™ Myriad™ 2 ?

Is there a figure about the savings or document to explain the difference?  

Thanks~

Backend  current support   future support

...

Intel Movidius™ Myriad™ 2 VPUs  coming soon,  yes

...

[Lu__Gavin]

Good question. I have the same feeling that part of functions of nGraph on optimization seems overlapped with OpenVINO, but not sure.

[Sreedhar_V_Intel]

Intel nGraph Objective is  focused on  training  phase of Deep Learning algorithms.

Consolidate hardware-independent and hardware-dependent optimizations for the benefit of all supported frameworks

Enable Intel training hardware to be utilized seamlessly by customers.

The model optimizer in OpenVINO tool kit takes pre trained models and convert into IR which can be used by Inference engine which is part of OpenVINO tool kit.

Thanks

Sreedhar