Home

Awesome

USD in a Docker Container

In this repository you will find a set of docker build scripts which will download all the Open Source packages required to build Pixar's Universal Scene Description, as well as USD itself.

If you are running a linux supported by nvidia-docker you can also run usdview interactively using the run.sh script, including CUDA acceleration for OpenSubdiv. Otherwise you can still use most tools such as usdcat.

The image tagged "usd-docker/vfx:centos7-usd-latest" will be very similar to a VFX Platform 2016 except for the Qt version still being at 4.8 as USD doesn't support Qt-5 and PySide-2 just yet.

The image tagged "usd-docker/usd:centos7-usd-latest" is built from the vfx one and adds USD.

We will not release pre-built images on Docker Hub for now as they are big and will contain dozens of OSS packages with very different licenses.

To build locally:

cd linux
./build-centos7.sh

To run usdview once built:

wget https://developer.apple.com/arkit/gallery/models/retrotv/retrotv.usdz --output-document=../data/retrotv.usdz
./run-centos7.sh usdview /data/retrotv.usdz

To build with maya

cd linux
./build-centos7_maya.sh 2018

Roadmap

At this stage this repository is used to test our AL_USDMaya plugin in a Animal Logic independent way and we are sharing it because it might help others when building USD.

This repository can also help evaluate USD as it makes all USD tools available within the docker container, these include usdview (with full GL and CUDA acceleration) and usdcat with Alembic support.

Goals

The main goal of this work is to ensure correct builds and automated tests of our AL_USDMaya plugin, as well as being a technology exploration, it is good to see the progress done by docker and nvidia on supporting GPU acceleration in containers.

One potential future goal could be to provide standard ways to build VFX Platform packages.

Build requirements

For easiest build you need a recent version of linux with Docker-1.9 and nvidia-docker. If you are on a platform not supported by docker-1.9 (such as CentOS-6) you can still build the images but running them is a lot trickier as nvidia-docker is not available and you have to manually create the volume containing the NVidia driver.

The builds have been tested on CentOS-6 and Ubuntu-14-10.

Maya images

We provide a few Dockerfiles that rely on having a maya devkit tarball available. Unfortunately the recent devkit are not accessible without Autodesk accounts which means we cannot provide automatic downloads for these.

We have tried to simplify the process by allowing you to download the tarballs and place them in a downloads folder and they will be fed to the build via a temporary python http server... this is not for the feint of heart. We would like to improve this in the future by "configuring" the Dockerfiles for specific targets depending on which builds are required.

Credits: