openvino-workshop-en

Overview

Hands-on workshop contents to learn Intel distribution of OpenVINO toolkit - a deep learning inferencing library. The workshop contents are tested on Ubuntu and Windows 10 systems.

Description

Intel distribution of OpenVINO toolkit is a library suite for computer vision applications. OpenVINO consists of following libraries and tools.

• Inference Engine - Efficient, high-performance and flexible deep learning inference run-time engine library
• Model Optimizer - Convert generic deep-learning models into OpenVINO IR format
• Model Downloader - Download OMZ (Open Model Zoo, Intel) models and popular deep learning models
• Deep Learning Workbench - Post training model re-quantization, benchmarking, accuracy checking
• OpenCV - High performance and feature-rich image processing library

OpenVINO provides great scalability. It supports wide variety of deep learning processors and accelerators. You can use almost the same code on different hardware easily.

• CPU - Atom to Xeon, OpenVINO supports the latest DL boot instructions
• Integrated GPU - OpenVINO can leverage the performance of integrated GPU and off load the task from CPU
• VPU - Vision Processing Unit (Myriad). A low power yet powerful deep-learning accelerator from Intel
• FPGA - OpenVINO compatible FPGA acclerator cards are available
• HDDL - High Density Deep Learning accelerator. Single or multiple Myriad devices are on a board

Also, OpenVINO supports various operating systems.

• Windows 10, Ubuntu, CentOS, MacOS

You will learn the basics of OpenVINO through this workshop.

1. Learning basic of OpenVINO API through a simple image classification program - classification.ipynb
2. Basic of object detection program using OpenVINO - object-detection-ssd.ipynb
3. Basic of asynchronous inferencing - classification-async-single.ipynb
4. Technique for high performance inference program - asynchronous and simultaneous inferencing - classification-async-multi.ipynb
5. < Appendix > Automate evaluation work on DevCloud - automated-testing.ipynb

How to use

1. Go to Intel distribution of OpenVINO toolkit web page and download an OpenVINO package suitable for your operating system
2. Install OpenVINO and setup support tools and accelerators by following the instruction in 'Get Started' page
3. Clone repository to your system

$ git clone https://github.com/yas-sim/openvino-workshop-en

4. Open a command terminal
5. Set up environment variables for OpenVINO

Linux $ source /opt/intel/openvino/bin/setupvars.sh  

Windows > call "Program Files (x86)\IntelSWTools\OpenVINO\bin\setupvars.bat"

6. Start Jupyter notebook
7. Open openvino-workshop-en/START-HERE.ipynb from Jupyter to start the workshop

Requirement

This workshop requires Intel distribution of OpenVINO toolkit.　Tested with OpenVINO 2020.1 version.

Contribution

Licence

Apache2

Author

Yasunori Shimura