zetton-docker

Docker images for development and deployment environment of Project Zetton.

Prebuilt images

Prebuilt images are shared on Docker Hub.

Building from source

The container image scripts are archived in the dist/ directory and are available for supported distros and package versions.

Here is an example on how to build an multi-arch container image based on vanilla Ubuntu 18.04:

./build.sh --package-name vanilla --os ubuntu --os-version 18.04 --arch amd64,arm64 --push

More examples:

# build vanilla images
./build.sh --package-name vanilla --os ubuntu --os-version 18.04 --arch amd64,arm64 --push
# build ROS images
./build.sh --package-name ros --package-version melodic --os ubuntu --os-version 18.04 --arch amd64,arm64 --push
# build CUDA images
./build.sh --package-name cuda --package-version 11.6.2 --os ubuntu --os-version 18.04 --arch amd64 --push
# build TensorRT images
./build.sh --package-name tensorrt --package-version 8.4.3.1 --os ubuntu --os-version 18.04 --arch amd64 --push --extra-args "--build-arg BASE_PACKAGE_VERSION=11.6.2"
# build ROS images with CUDA
./build.sh --package-name ros --package-version melodic  --os ubuntu --os-version 18.04 --arch amd64 --push --image-suffix cuda11.6.2 --extra-args "--build-arg BASE_PACKAGE=cuda-11.6.2"
# build ROS images with TensorRT
./build.sh --package-name ros --package-version melodic  --os ubuntu --os-version 18.04 --arch amd64 --push --image-suffix tensorrt8.4.3.1 --extra-args "--build-arg BASE_PACKAGE=tensorrt-8.4.3.1"

See ./build.sh --help for detailed usage.

Usage

Image Tags

projectzetton/zetton-docker:<package_name>(-<package_version>)-<category>-<os_version>

• <package_name>:
  • vanilla
  • ros
  • cuda
  • tensorrt
• <package_version> (optional)
  • vanilla is empty
  • ros is the ROS distro name
  • cuda is the CUDA version
  • tensorrt is the TensorRT version
• <category>:
  • base
  • runtime
  • devel
• <os_version>:
  • ubuntu18.04

Packages

• vanilla: pure Ubuntu image with basic tools
• ros: vanilla image with ROS installed
• cuda: Ubuntu image with CUDA installed
• tensorrt: cuda image with TensorRT and othter deep learning related packages (e.g., PyTorch, TensorFlow and ONNX) installed

Categories

The base image provides a minimal environment with the following modifications:

• optional APT mirror
• correct timezone

The runtime image provides environments which contains the following packages:

• CMake
• OpenCV
• absl
• fmt

In addition to the packages included in the runtime image, the ``devel` image also contains the following features.

• SSH on port 2222 with password
• yapf & clang-format & clangd
• Zsh with Oh-My-Zsh
• Custom Tmux configuration

Acknowledgement

• The container image scripts are based on nvidia/container-images/cuda