You can easily get started by using the precompiled and published docker images. In order to start using them you need to meet the following prerequisites:

• Linux operating system or Windows Subsystem for Linux (WSL2)
• Installed docker engine or compatible container engine
• Permissions to start containers (sudo or docker group membership)

Pull a docker image

docker pull openvino/ubuntu20_dev:latest

Start the container with an interactive session

export IMAGE=openvino/ubuntu20_dev:latest
docker run -it --rm $IMAGE /bin/bash

Inside the interactive session, you can run all OpenVINO samples and tools.

Run a python sample

If you want to try samples, then run the image with the command like below:

docker run -it --rm $IMAGE /bin/bash -c "python3 samples/python/hello_query_device/hello_query_device.py"

Download a model via omz_downloader

docker run -it -u $(id -u):$(id -g) -v $(pwd)/:/model/ --rm $IMAGE \
/bin/bash -c "omz_downloader --name googlenet-v1 --precisions FP32 -o /model"

Convert the model to IR format

docker run -it -u $(id -u):$(id -g) -v $(pwd)/:/model/ --rm $IMAGE \
/bin/bash -c "omz_converter --name googlenet-v1 --precision FP32 -d /model -o /model"

In result, the converted model will be copied to public/googlenet-v1/FP32 folder in the current directly:

tree public/googlenet-v1/
public/googlenet-v1/
├── FP32
│   ├── googlenet-v1.bin
│   └── googlenet-v1.xml
├── googlenet-v1.caffemodel
├── googlenet-v1.prototxt
└── googlenet-v1.prototxt.orig

Run a benchmark app

docker run -it -u $(id -u):$(id -g) -v $(pwd)/:/model/ --rm $IMAGE benchmark_app -m /model/public/googlenet-v1/FP32/googlenet-v1.xml

Check also:

Prebuilt images

Working with OpenVINO Containers

Deployment with GPU accelerator

Generating dockerfiles and building the images in Docker_CI tools