Docker

Viseron is distributed as a Docker image. This makes deployment easy for the user and reduces the "But it works on my machine" factor.

Develop Docker containers

Viseron heavily uses multistage Docker builds, and compilation of different components are split into different Dockerfiles.

Rebuild Viseron image

If you have made changes to the Viseron code and want to rebuild the image, you can run the following command.

docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-viseron

tip

If you are developing on a different architecture than amd64 or you have a CUDA capable GPU, you can replace amd64 with the architecture you are developing on.

See the table below for supported architectures.

Rebuild Python dependencies (wheels)

In order to speed up builds, Viseron uses precompiled Python wheels.

If your changes requires a new or updated Python dependency, you have to rebuild the wheels image.

• Edit requirements.txt in the root of the repository (preferably by pinning the version of the dependency).
• Build the wheels image:

  docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-wheels
• Build Viseron image:

  docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-viseron

Upgrading library versions

Lets say you want to update to a newer version of OpenCV. To do this you would:

• Edit OPENCV_VERSION in ./azure-pipelines/.env
• Build the OpenCV image:

  docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-opencv
• Build Viseron image:

  docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-viseron

Rebuild all images

To build all the amd64 image from scratch the following commands can be used.

docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-ffmpeg && \
docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-opencv && \
docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-dlib && \
docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-wheels && \
docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-base && \
docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build amd64-viseron

Cross-building

To support different CPU architectures i use Balenas base images that runs QEMU. To build these you need to first register QEMU on your builder.

The easiest way to do that is to run:

docker run --rm --privileged tonistiigi/binfmt --install all

You can then simply build the containers like you normally would and QEMU will be invoked automatically.

docker compose --file ./azure-pipelines/docker-compose-build.yaml --env-file ./azure-pipelines/.env build aarch64-viseron

Architecture and Hardware specific images

The images are prefixed with the architecture or hardware they are built for.

Prefix	Description
aarch64	Generic aarch64 image
amd64	Generic image
amd64-cuda	CUDA enabled images
rpi3	RaspberryPi 3
jetson-nano	Jetson Nano CUDA support