Develop Multiplatform Computer Vision Solutions with Intel® Distribution of OpenVINO™ Toolkit

Print Friendly, PDF & Email

Sponsored Post

Realize your computer vision deployment needs on Intel® platforms—from smart cameras and video surveillance to robotics, transportation, and much more. The Intel® Distribution of OpenVINO™ Toolkit (includes the Intel® Deep Learning Deployment Toolkit) allows for the development of deep learning inference solutions for multiple platforms. The toolkit allows developers to build applications that emulate human vision, and is based on convolutional neural networks (CNNs), extending workloads across Intel® hardware (including accelerators) and maximizes performance in the following ways:

  • Enables deep learning inference from edge to cloud.
  • Supports heterogeneous execution across Intel platforms and accelerators—CPU, GPU, VPU, and FPGA—using a common unified API.
  • Speeds up time to market via a library of functions and pre-optimized kernels.
  • Provides extensibility and supports custom layer implementations.
  • Includes the Deep Learning Workbench, a GUI tool for running inference experiments and determine optimal configurations.
  • Includes optimized calls for OpenCV and OpenVX.
Computer vision capabilities even at the edge with OpenVINO™

Improved Neural Network Performance

The OpenVINO™ toolkit provides developers with improved neural network performance on a variety of Intel® processors and helps further unlock cost-effective, real-time AI applications. The toolkit enables deep learning inference and straightforward heterogeneous execution across multiple Intel® platforms (CPU, Intel® Processor Graphics)—providing implementations across cloud architectures to edge devices.  

The OpenVINO™ toolkit is an open-source product. It contains the Deep Learning Deployment Toolkit (DLDT) for Intel® processors (for CPUs), Intel® Processor Graphics (for GPUs), and heterogeneous support. It also includes an open model zoo with pre-trained models, samples, and demos.

Optimized models use the Deep Learning Deployment Toolkit from Intel and the Intel® Math Kernel Library for Deep Neural Networks (Intel® MKL-DNN) to deliver outstanding inferencing performance for practical deployment of AI solutions at the “edge” of the enterprise in clinical and research settings

Reference Implementations

There are a number of compelling reference implementation for deploying OpenVINO™. These implementations afford data scientists and developers a significant head-start for developing solutions in a number of different problem domains:

  • Vehicle detection and license plate recognition.
  • People counter systems – this reference implementation is designed to detect people in a designated area and determine the number of people in the frame, the average time they are in the frame, and the total count.
  • Shopper gaze monitor – this application uses the Inference Engine included in OpenVINO™ and the Intel® Deep Learning Deployment Toolkit. It is designed for a retail shelf mounted camera system that counts the number of passers-by that look towards the display vs. the number of people that pass by the display without looking.
  • Digital kiosk display for 4K ads – this application identifies the age and gender of the audience standing in front of digital signage, and based on the identification, it selects a suitable 4K advertisement.
  • Store isle monitor – this reference implementation counts the number of people present in an image and generates a motion heatmap. It takes the input from the camera, or a video file for processing. Snapshots of the output are taken at regular intervals and are uploaded to the cloud. It also stores the snapshots of the output locally.
  • Store traffic monitor – this application is one of a series of IoT reference implementations aimed at instructing data scientists and developers on how to develop a working solution for a particular problem. It demonstrates how to create a smart video IoT solution using Intel® hardware and software tools. This reference implementation monitors people activity inside and outside a facility, as well as counting product inventory.
  • Safety gear detector – this application is an IoT reference implementation that detects people and potential violations of safety-gear standards.
  • Restricted zone notifier – this application uses the Inference Engine included in the Intel® Distribution of OpenVINO™ toolkit and the Intel® Deep Learning Deployment Toolkit. A trained neural network detects people within a marked assembly area, which is designed for a machine mounted camera system. It sends an alert if there is at least one person detected in the marked assembly area.
  • Object size detection – this application demonstrates how to use computer vision to detect and measure the approximate length, width and size of assembly line parts. It is designed to work with an assembly line camera mounted above the assembly line belt. The application monitors mechanical parts as they are moving down the assembly line and raises an alert if it detects a part on the belt outside a specified size range.
  • Intruder detector – this application is an IoT reference implementation aimed at demonstrating how to create a smart video IoT solution using Intel® hardware and software tools. This solution detects any number of objects in a designated area, providing the number of objects in the frame and total count.

A Strategic Toolkit for Data Scientists

The OpenVINO™ toolkit enables CNN-based deep learning inference on the edge for computer vision applications. The toolkit is specifically designed for data scientists and software developers who work on AI applications, computer vision, neural network inference, and deep learning deployment capabilities. It is also for those who need to accelerate their solutions across multiple platforms including CPU, GPU, VPU, and FPGA.

Speak Your Mind



  1. Is it possible to extend OpenVINO to work also on AMD GPU devices?