Universal Scene Description in Docker

This project aims to build an easy to use Docker container that contains everything needed for me to experiment with Pixar's Universal Scene Description (USD) in a safe and reproducible environment.

Currently, the container is built on top of NVIDIA's pre-built USD packages, targeting USD 21.5 and Python 3.6. These versions can be changed by cloning this repository, editing the Dockerfile, and building the container yourself.

Basic usage

The entire USD toolset is available, and on the $PATH inside the container (handled by the entrypoint script automatically). This means that you can simply run one of the pre-built USD commands, such as usdview, usdcat, or usdwrite, and it will be run inside the container. For example, we can cat one of the example USD files in the container:

docker run --rm ewpratten/usd usdcat /usr/local/nvidia/usd/share/usd/tutorials/convertingLayerFormats/Sphere.usd

Note: Inside the container, USD is installed in /usr/local/nvidia/usd. This location is stored in the environment variable $USDROOT, accessible from any scripts run in the container.

X11 passthrough

Some USD utilities (for example, usdview) require an OpenGL context, and a display to draw to. This is usually not possible in a regular Docker container (at least not without a lot of extra work). To get around this, the USD Docker container is built on NVIDIA's nvidia/opengl:base-ubuntu20.04 base image, and due to which, can be run via the NVIDIA Container Toolkit. To view the example HelloWorld.usd file used in Pixar's Referencing Layers tutorial, we can run the following command (on a Linux host):

nvidia-docker run -it \
    --user=$(id -u $USER):$(id -g $USER) \
    --env DISPLAY \
    -v /tmp/.X11-unix:/tmp/.X11-unix \
    --env QT_X11_NO_MITSHM=1 \
    ewpratten/usd \
    usdview /usr/local/nvidia/usd/share/usd/tutorials/referencingLayers/HelloWorld.usda

Running your own code

Aside from a few modifications to the environment and package versions, this is essentially just a regular Ubuntu 20.04 Docker base image, and can be used as such.

To execute your own code in the container, either extend the image with our own Dockerfile, or mount your application as a volume to ewpratten/usd. Keep in mind, when doing either, that the USD Docker container uses a custom entrypoint.sh script, which does a bit of environment setup. Make sure to initialize the environment yourself if overriding the container entrypoint.

VSCode Devcontainers

In order to get autocomplete for the pxr Python library in VSCode without building USD on my development machine, I have built a custom devcontainer for USD development. Using it in your editor is as simple as creating a file at .devcontainer/devcontainer.json with the following contents:

{
    "name": "USD Development",
    "image": "ewpratten/usd-devcontainer:latest",
    "extensions": [
        "ms-python.vscode-pylance",
        "ms-python.python"
    ],
    "remoteUser": "vscode"
}