Five Easy Steps to Deploy the Intel® Distribution of OpenVINO™ Toolkit

This article is for ISVs with existing Computer Vision and AI-centric applications at the edge who would like to optimize their solutions for Intel® hardware. It takes advantage of the many features you can find in the Intel® Distribution of OpenVINO™ toolkit.

If one of the below scenarios accurately describes your current development approach, this article provides a well-defined path to augment your Computer Vision Solution:

  • Currently using deep learning models like Tensorflow* and Caffé* 

  • Currently using traditional Computer Vision with the potential to start using deep learning

The options available for developing with Computer Vision have grown. We encourage you to find the Intel® tools and hardware to take your application to the next level.

Here are 5 easy steps that will take you from prototyping to deployment, and finally to scale, while opening doors for cross-market collaboration opportunities through complementary Intel® ecosystem programs.

STEP 1: Understanding How Intel® Distribution of OpenVINO™ Toolkit Can Help Enhance Your Computer Vision Application 

For your computer vision use case, Intel provides various open source technologies to suit your developer needs.  Whether you’re starting from deep learning or traditional computer vision development, Intel provides the tools and resources to improve your application performance and power efficiency at little or no cost.  We have a variety of tools to optimize your existing models in the Intel® Distribution of OpenVINO™ toolkit as well as developer kits and accelerators from the industrial level IEI Tank* AIoT Developer Kits to smaller form factor, low-cost accelerators, such as the Intel® Neural Compute Stick to help complete the development of your vision solution. 

Your Programming Model

How to leverage the Intel® Distribution of OpenVINO™ toolkit?

Using Traditional Computer Vision

CV Libraries for improved performance

Pre-trained models, so you don’t need to create your models from scratch

Using Deep Learning Models like Tensorflow* and Caffé*

Import Models from various supported frameworks - Caffe*, TensorFlow*, MXNet*, ONNX*, Kaldi*

100+ models for Caffe, MXNet and TensorFlow validated

Explore the features of the Intel® Distribution of OpenVINO™ toolkit, and the new programming model, to deploy applications on a range of Intel® silicon in this 8 Part Video Series.

STEP 2: Optimize your Solution Using the Intel® Distribution of OpenVINO™ Toolkit

This tool includes a selection of pre-optimized models, such as object detection and facial demographics for advanced video analytics. In addition, learn how to optimize your own model using the Model Optimizer and take advantage of heterogeneous compute with the Inference Engine.

Setup the Intel® Distribution of OpenVINO™ toolkit on your development box:

Setup with Accelerators:

Optimize your Models:

  • Model Optimizer:  a cross-platform tool that adjusts deep learning models for optimal execution on end-point target devices

Use Pre-optimized models:

  • Optimized Models: help expedite development and improve image processing pipelines for Intel® processors

STEP 3: Accelerate Prototyping with Modular Reference Implementations & Code Samples

Select from numerous open source reference implementations to help kick start your development. Deploy your own IoT solutions by using these flexible, open source, and modular code samples. Integrate reference implementations with your existing solutions, or use them to enhance your solution to add new compatibilities. Learn how to unleash the power of our Intel® Distribution of OpenVINO™ toolkit tool.

STEP 4: Identify Hardware Options Best Suited for Your Solution

Intel offers a wide range of hardware options for your unique project requirements and industry vertical. If you are just getting started with your IoT Solution, UP Squared* AI Vision X Development Kit is the right choice for your prototyping journey. Looking for robust performance and scalable options, then choose the IEI Tank* AIoT Developer Kit with an Intel® Core™ processor or an Intel® Xeon® processor.  These developer kits come with pre-installed developer tools and SDKs, enable quick setup with preloaded samples, and offer performance to support multiple workloads.

In addition, Intel® Vision Accelerator Design products — Intel® Movidius™ VPU and Intel® Arria® 10 FPGA — allow you to take advantage of powerful, deep neural network inference for fast, accurate video analytics. All hardware developer kits come with pre-installed software and a choice of accelerators supporting CPU, GPU, VPU, and FPGA—using a common API. Developer kits are also compatible with Intel® Movidius™ Neural Compute Stick 2 for small form factor solutions and fast prototyping.

Choose the hardware and accelerators that are best suited to your needs, from prototyping to production.

Intel Hardware

Use Case

What’s Included

UP Squared* AI Vision X Development Kit 


  • Small form factor for small workload fitting into smaller spaces
  • Low power
  • Includes an HD USB camera for vision applications
  • Includes CPU, integrated GPU, and VPU for heterogeneous compute and balance between flexibility and performance per watt across processing units


  • Ubuntu 16.04 with preinstalled dev tools and SDKs
  • Intel® Distribution of OpenVINO™ toolkit, packaged with optimized  pretrained models
  • Intel® System Studio - all in one cross-platform app building tool


IEI Tank* AIoT Developer Kit, Intel® Core™ processor 

IEI Tank* AIoT Developer Kit, Intel® Xeon® processor   


  • Fanless design and ruggedized case for industrial deployments
  • Handle multiple video streams for running as an AI processing hub


  • Ubuntu 16.04 with preinstalled dev tools and SDKs
  • Intel® Distribution of OpenVINO™ toolkit, packaged with optimized  pretrained models
  • Intel® System Studio - all in one cross-platform app building tool


Intel® Vision Accelerator Design Products

Intel® Vision Accelerator Design product bundles to match the performance, cost, and power-efficiency required on various Intel® architectures utilizing the Intel® Core®™ processor and Intel® Xeon® processor


Eight Intel® Movidius™ Myriad™ X Vision Processing Unit (VPU) 2485: Typically supports 1 to 16 video streams per
device (depends on desired frame rate and
algorithm complexity)


One Intel® Arria® 10 FPGA: Aggregation of > 32 streams of video, speech, and other sensor data


Intel® Neural Compute Stick  2


  • Prototype Development Platform for computer vision
  • High-performance vision processing unit: Intel® Movidius™ Myriad™ X VPU ideal for AI applications



Apply multiple deep neural networks (DNN) on multiple sticks for optimized performance

STEP 5: Scale Through Intel® Ecosystem Programs & Joint Development Opportunities

Once you are ready with your solution, explore the cross-market collaboration opportunities available through Intel® ecosystem programs. AI: In Production Program: aims to accelerate the development of AI-centric applications at the edge. Become a member and gain early access to tools and technologies, increased exposure with amplified marketing, exclusive networking and matchmaking, and scalability through the Intel® ecosystem. Discover joint development opportunities and increase cross-market collaboration.

Check out the other complementary Intel® ecosystem programs

For more complete information about compiler optimizations, see our Optimization Notice.