Open3D in Docker and VS Code

Using the Open3D library inside Docker and VS Code sounds like an easy task, but don't be fooled. I made this repo so that you don't have to suffer like I did. There were many hurdles.Thanks to this issue I figured out that I needed to use the nvidia/cudagl images as my base.

There are other OpenGL enabled images out there which also might do the trick, but I require CUDA for future projects which is why I chose that one.

Prerequisites

• Remote Development extension

Installation / Building

git clone https://github.com/SebastianGrans/Open3DinDocker.git
cd Open3DinDocker
chmod +x makeJSONfile.sh 
# Updates the VS Code devcontainer.json file with your username and 
./makeJSONfile.sh 

Open the folder in VS Code

code .

Press Ctrl+⇧ Shift+P to open the commande palette and execute the command Remote-Container: Reopen in container. This will build and image and start a container.

Running Open3D

Run the file testOpen3D.py. Either via VS Code (F5) or from the terminal: python testOpen3D.py.

Note: When you run this the first time after starting the container, it might take quite a while to import Open3D (~30 sec). Executing python -vc "import open3d" indicates that it get stuck at import 'open3d._build_config' for a while before rapidly continuing.

If everything worked you should see the following:

Running xeyes

The most basic test. In the integrated terminal, run xeyes and you should be greeted with:

Running glxgears

glxgears is a classic test for checking that OpenGL is working.

Running Google Chrome / Chromium

I'm not sure why you would do this, but you can in fact run Chrome from inside a Docker container.

Chromium

sudo apt install chromium-browser

Then run it:

chromium-browser --no-sandbox

Google Chrome

sudo apt install wget
wget https://dl.google.com/linux/direct/google-chrome-stable_current_amd64.deb
sudo apt install ./google-chrome-stable_current_amd64.deb

Then run it:

google-chrome --no-sandbox

Note: If you want to run it without --no-sandbox you can supposedly run the container with the --privileged flag.

How does it work?

The command for making X11 forwarding is quite easy and I found it on the ROS wiki.

You essentially mount all relevant user information files from the host machine into the container and run it using your default user, rather than root which is the default. Since you as a user already has permission to the X11 server, it is also permitted from inside the container.

Building and running in a terminal is easier.

Building
Inside this repo, run:

docker build -t x11docker .

This build a new image called x11docker.
Run it:

docker run -it \
    --user=$(id -u $USER):$(id -g $USER) \
    --env="DISPLAY" \
    --volume="/etc/group:/etc/group:ro" \
    --volume="/etc/passwd:/etc/passwd:ro" \
    --volume="/etc/shadow:/etc/shadow:ro" \
    --volume="/etc/sudoers.d:/etc/sudoers.d:ro" \
    --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" \
    --volume="$(pwd):/home/${USER}/" \
    --workdir="/home/${USER}/" x11docker

And you should be greeted by a bash promt.

Sources that helped me along the way:

• General X11 forwarding in Docker http://wiki.ros.org/docker/Tutorials/GUI
• CUDA Docker image with OpenGL NVIDIA/nvidia-docker#534
• ZSH on the host machine causes problems when mounting microsoft/vscode-remote-release#129 (comment)
• Shell commands are not evaluated in the devcontainer.json-file microsoft/vscode-remote-release#551