Community
cancel
Showing results for 
Search instead for 
Did you mean: 
dilip96
New Contributor I
368 Views

How compatible is Myriad 2 or X for developing a product? The application uses a custom trained TensorFlow SSD MobileNetV2 model for object detection along with inference performed by the Open CV DNN module.

I am planning to buy a Myriad 2 or X and start working on them as an edge inference device. Please detail me on the Mentioned technologies compatibility (especially OpenCV DNN module). And also a brief insight on how to get started with the product will also be helpful, since I am very new to work with Intel boards.

 

And also, does the product accelerate only Inference of deep learning models are other image processing techniques too? In the application some image processing algorithms are run after the inference. Does the product give me some acceleration in those function?

 

Please provide me as much as details possible on these. Thank you.

0 Kudos
7 Replies
MAURICIOAL_R_Intel
Moderator
182 Views

Thank you for reaching out.

 

For the DNN module, we recommend you to take a look at its documentation here. Myriad is listed as a target device, so it should work with the Intel® Neural Compute Stick 2 (NCS2), you can check this guide for more information.

 

Does the product accelerate only the Inference of deep learning models? Are other image processing techniques too?

The Intel® Neural Compute Stick 2 only accelerates the inference of deep learning models. You can use OpenVINO Toolkit to optimize models for the Intel® Neural Compute Stick 2.

 

Also, you can check the Intel® Neural Compute Stick 2 starting guide here.

 

Regards,

Mauricio R.

dilip96
New Contributor I
182 Views

Hi Mauricio,

 

Thanks for your immediate assistance. Information provided is very helpful.

 

Myriad is listed as a target device, so it should work with the Intel® Neural Compute Stick 2 (NCS2).

I would like to have some clarification on this. I assumed NCS2 and Myriad was two different devices, but looking similarly.

Is it that, NCS2 is the hardware which is available with Myriad 2 and Myriad x variants?

 

Also I would like to know if there is any Development boards available on this, so that I can get familiarized working on them.

 

Thank You.

MAURICIOAL_R_Intel
Moderator
182 Views

Thank you for your response.

 

To clarify your question, the Intel® Neural Compute Stick 2 (Intel® NCS 2) is built with exceptional performance on the Intel® Movidius™ Myriad™ X Vision Processing Unit (VPU).

 

There are some development boards with the Intel Movidius Myriad X VPU, like the UP AI Core X. Also, you can check more information here. But if you want to use the Intel® Neural Compute Stick 2 as a USB form factor with Myriad X, you can use it on a Linux, Windows, macOS or even Raspberry Pi.

 

Regards,

Mauricio R.

dilip96
New Contributor I
182 Views

HI, very much happy with your response.

 

Quikly summarizing my understanding,

 

The Myraid VPU is available in two products,

  • NCS2
  • AICore X with UP SOC boards - where different boards are available with number of VPU's (1,2,3,4,8 VPU's).

 

Please recommend me a method to start working . Since I have plan of Developing a product with this, which product will give me more understanding and make the future work easier.

 

Also, my algorithm runs a SSD MobileNet V2 , and some image processing techniques after the detection. So what is the Number of VPU's sufficient for my application.

 

Thank You.

MAURICIOAL_R_Intel
Moderator
182 Views

Thank you, Dilip.

 

We support the ssd_mobilenet_v2_coco model here: http://download.tensorflow.org/models/object_detection/ssd_mobilenet_v2_coco_2018_03_29.tar.gz.

 

We recommend you prototyping on the Intel® Neural Compute Stick 2 as it has a low price point. For your interest, Intel® Neural Compute Stick 2 is not for image processing, it is just for inference.

 

For additional devices with Intel Myriad VPUs, we recommend you to check Intel Vision Accelerator Design with Intel Movidius VPUs. You will need to decide which device adapts better to your project.

 

Regards,

Mauricio R.

dilip96
New Contributor I
182 Views

​Thank you so much for your support. I have decided to start with Intel® Neural Compute Stick 2.

This conversation was very helpful.

 

 

MAURICIOAL_R_Intel
Moderator
182 Views

Thank you for your response.

 

I'm glad that this helped you.

 

Regards,

Mauricio R.