OpenVINO Docker

Collection of Dockerfiles that will provide you with a base environment to build and run your inference models with Intel® OpenVINO™ Toolkit

OpenVINO can be installed downloading the installation files from the official web, using the Docker Hub images, using YUM or APT packages.

These Dockerfiles uses the APT package on top of Ubuntu 18.04

Building the Docker Images

docker-compose build

This will build the following images:

REPOSITORY                          TAG                 IMAGE ID            CREATED             SIZE
sampleapp-runtime                   latest              199338615d9e        2 minutes ago       375MB
sampleapp-dev                       latest              99298ab0a7e3        3 minutes ago       1.74GB
openvino-dev                        2020.3.194          e1462a646c16        4 minutes ago       710MB
openvino-runtime                    2020.3.194          f9b55d3ad15e        5 minutes ago       562MB
ubuntu                              18.04               c3c304cb4f22        3 weeks ago         64.2MB

ubuntu: is the base image
openvino-runtime: is the runtime OpenVINO image based on APT package intel-openvino-runtime-ubuntu18-2020 3.194. Includes the following modules:

  Depends: intel-openvino-docs-2020.3.194
  Depends: intel-openvino-eula-2020.3.194
  Depends: intel-openvino-gstreamer-rt-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-gva-rt-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-rt-2020.3.194
  Depends: intel-openvino-ie-rt-core-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-rt-cpu-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-rt-gna-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-rt-gpu-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-rt-hddl-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-rt-vpu-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-sdk-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-opencv-generic-2020.3.194
  Depends: intel-openvino-opencv-lib-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-setupvars-2020.3.194

openvino-dev: is the development OpenVINO image including all the components and based on APT package intel-openvino-dev-ubuntu18-2020.3.194. Inludes the openvino-runtime modules with the addition of:

  Depends: intel-openvino-dl-workbench-2020.3.194
  Depends: intel-openvino-gva-dev-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-gva-sdk-2020.3.194
  Depends: intel-openvino-ie-bin-python-tools-ubuntu-bionic-2020.3.194
  Depends: intel-openvino-ie-samples-2020.3.194
  Depends: intel-openvino-model-optimizer-2020.3.194
  Depends: intel-openvino-omz-dev-2020.3.194
  Depends: intel-openvino-omz-tools-2020.3.194
  Depends: intel-openvino-opencv-etc-2020.3.194
  Depends: intel-openvino-pot-2020.3.194

sampleapp: is a sample application using the OpenVINO Toolkig.

Using the image

Run a sample application

We will run the security barrier demo included with OpenVINO, in a separate container.

docker-compose run sampleapp-runtime

The inference output should be visible in the terminal:

$ docker-compose run sampleapp-runtime 
[setupvars.sh] OpenVINO environment initialized
[ INFO ] InferenceEngine: 0x7f8c44aaf030
[ INFO ] Files were added: 1
[ INFO ]     /opt/intel/openvino/deployment_tools/demo/car_1.bmp
[ INFO ] Loading device CPU
	CPU
	MKLDNNPlugin version ......... 2.1
	Build ........... 42025

[ INFO ] Loading detection model to the CPU plugin
[ INFO ] Loading Vehicle Attribs model to the CPU plugin
[ INFO ] Loading Licence Plate Recognition (LPR) model to the CPU plugin
[ INFO ] Number of InferRequests: 1 (detection), 3 (classification), 3 (recognition)
[ INFO ] 4 streams for CPU
[ INFO ] Display resolution: 1920x1080
[ INFO ] Number of allocated frames: 3
[ INFO ] Resizable input with support of ROI crop and auto resize is disabled
Invalid MIT-MAGIC-COOKIE-1 keyUnable to init server: Could not connect: Connection refused

(Detection results:13): Gtk-WARNING **: 08:35:06.722: cannot open display: :0

Run the the container with X enabled (Linux)

This sample uses OpenCV to desplay the resulting frame with detections rendered as bounding boxes and text. For running a sample application that displays an image, you need to share the host display to be accessed from guest Docker container.

First the X server on the host should be enabled for remote connections (note that this turns off access control):

xhost +

The following flags needs to be added to the docker run command:

--net=host
--env="DISPLAY"
--volume="$HOME/.Xauthority:/root/.Xauthority:rw"

This is already added in the docker-compose.yml:

    volumes:
      - $HOME/.Xauthority:/root/.Xauthority
    network_mode: host
    environment:
      - DISPLAY

Now, run again the sample app and you will see the screen output

When finished, disable the remote connections to the X server

xhost -

Use the image in another container

You can use this Docker image as a base image and use it in multiple Dockerfiles. Use multi-stage with openvino-dev and/or openvino-runtime image in your Dockerfile.

In general, openvino-dev would be used when using:

Deep Learning Workbench
Gstreamer Video Analytics development and SDK
Infrence Engine Python tools and Samples
Model Optimizer
OpenVINO Model Zoo Tools
Intel® Post Training Optimization Tool

Otherwise, openvino-runtime may be used

## Development build
FROM openvino-dev:2020.3.194 as sampleapp-dev

RUN <Application build steps, Model Download, Model Optimization, etc>
RUN <Deployment Tool>

##  Runtime Build

FROM ubuntu:18.04

COPY --from=sampleapp-dev <Application Biuld>
COPY --from=sampleapp-dev <Models and data>

CMD ["/myapp"]

An example of this process is shown in sample-app folder

mateoguzman / openvino-docker