dockerfiles

Dockerfiles with Intel® Distribution of OpenVINO™ toolkit

This repository folder contains Dockerfiles to build an docker image with the Intel® Distribution of OpenVINO™ toolkit. You can use Docker CI framework to build an image, please follow Get Started with DockerHub CI for Intel® Distribution of OpenVINO™ toolkit.

1. Supported Operating Systems for Docker image
2. Supported devices and distributions
3. Where to get OpenVINO package
4. How to build
5. Prebuilt images
6. How to run a container

Supported Operating Systems for Docker image

• ubuntu18 folder (Ubuntu* 18.04 LTS)
• ubuntu20 folder (Ubuntu* 20.04 LTS)
• rhel8 folder (RHEL* 8)
• winserver2019 folder (Windows* Server Core base OS LTSC 2019)
• windows20h2 folder (Windows* OS 20H2)

Note: dl-workbench folder contains Dockerfiles for OpenVINO™ Deep Learning Workbench.

Supported devices and distributions

Devices:

• CPU
• GPU
• VPU (NCS2)
• HDDL (VPU HDDL) (Prerequisite: run HDDL daemon on the host machine, follow the configuration guide for HDDL device)

OpenVINO documentation for supported devices.

Distributions:

• runtime: IE core, nGraph, plugins
• dev: IE core, nGraph, plugins, samples, Python dev tools: Model Optimizer, Post training Optimization tool, Accuracy checker, Open Model Zoo tools (downloader, converter)
• base (only for CPU): IE core, nGraph

NOTE: OpenCV is not included since OpenVINO 2022.1.0. You can install it separately using the download script in the extras/scripts directory (download_opencv.sh for Linux and download_opencv.ps1 for Windows).

You can generate Dockerfile with your settings, please follow the DockerHub CI documentation.

• runtime and dev distributions are based on archive package of OpenVINO product. You can just remove unnecessary parts.
• base distribution is created by OpenVINO™ Deployment Manager.

Where to get OpenVINO package

You can get OpenVINO distribution packages (runtime, dev) directly from public storage. For example:

• take runtime l_openvino_toolkit_runtime_ubuntu20_p_2022.1.0.643.tgz package and specify -dist runtime option for Docker CI docker_openvino.py or take a Dockerfile with runtime suffix.
• take dev l_openvino_toolkit_dev_ubuntu20_p_2022.1.0.643.tgz package and specify -dist dev option for Docker CI docker_openvino.py or take a Dockerfile with dev suffix.

How to build

Note: Please use Docker CI framework release version corresponding to the version of OpenVINO™ Toolkit that you need to build.

• Base image with CPU only:

You can use Docker CI framework to build an image, please follow Get Started with DockerHub CI for Intel® Distribution of OpenVINO™ toolkit.

python3 docker_openvino.py build --file "dockerfiles/ubuntu18/openvino_c_base_2020.3.dockerfile" -os ubuntu18 -dist base -p 2020.3.341

Or via Docker Engine directly, but you need specify BUILD_ID argument:

docker build --build-arg BUILD_ID=2020.3.341 -t ubuntu18_base_cpu:2020.3.341 - < dockerfiles/ubuntu18/openvino_c_base_2020.3.dockerfile

---

• Dev/runtime image:

You can use Docker CI framework to build an image, please follow Get Started with DockerHub CI for Intel® Distribution of OpenVINO™ toolkit.

python3 docker_openvino.py build --file "dockerfiles/ubuntu18/openvino_cgvh_dev_2021.dockerfile" -os ubuntu18 -dist dev -p 2021.1

For runtime distribution, please set appropriate -dist and --file options.

Or via Docker Engine directly, but you need specify package_url argument (see Where to get OpenVINO package section):

docker build --build-arg package_url=https://storage.openvinotoolkit.org/repositories/openvino/packages/2021.4/l_openvino_toolkit_dev_ubuntu18_p_2021.4.582.tgz \
             -t ubuntu18_dev:2021.4 -f dockerfiles/ubuntu18/openvino_cgvh_dev_2021.4.dockerfile .

---

• Custom image with CPU, iGPU, VPU support
  You can use Dockerfiles from the build_custom folders to build a custom version of OpenVINO™ from source code for development. To learn more, follow:
  • Build custom Intel® Distribution of OpenVINO™ toolkit Docker image on Ubuntu 18
  • Build custom Intel® Distribution of OpenVINO™ toolkit Docker image on Ubuntu 20

Prebuilt images

Prebuilt images are available on:

• Docker Hub
• Red Hat* Quay.io
• Red Hat* Ecosystem Catalog (runtime image)
• Red Hat* Ecosystem Catalog (development image)
• Azure* Marketplace

How to run a container

Please follow Run a container section in DockerHub CI getting started guide.

Documentation

• Install Intel® Distribution of OpenVINO™ toolkit for Linux* from a Docker* Image
• Install Intel® Distribution of OpenVINO™ toolkit for Windows* from Docker* Image
• Official Dockerfile reference

---

* Other names and brands may be claimed as the property of others.