udacity-carnd

Creates a Docker image with all the prerequisites needed to run the projects of the Udacity Self-Driving Car Nanodegree.

This image is based on the dev-machine image, and extends it by adding Anaconda3, OpenCV, TensorFlow, Keras, uWebSockets, cppad, ipopt, etc as instructed in the nanodegree. I have used it to successfully complete all projects of the first and second terms.

This image can be launched in CLI mode, in UI mode via a TurboVNC server that you can connect to, or as a jupyter server. Or in a variety of other situations.

The most useful way to launch it, is probably as a jupyter server. The image has a useful script baked in, that you can run as:

$ docker run -it -v $PWD:/src -p 8888:8888 yrahal/udacity-carnd bash /bin/run_jupyter.sh

You can then open localhost:8888 in your browser and execute your notebooks, assuming your notebooks can be found under the current directory hierarchy.

You can also refer to the instructions from the dev-machine repository for extra information on usage. The notable differences is that yrahal/dev-machine must be replaced by yrahal/udacity-carnd and that the default user in yrahal/udacity-carnd is... kitt.

You might also find this Medium post and this YouTube video useful.

Files

• run.sh: Script provided for convenience to run the image with some useful mappings:
  • Runs the image with a TurboVNC server and maps the container's 5901 port to the same one on the host. It also maps the 8888 jupyter and 4567 simulator ports to that of the host.
  • Maps the current directory on the host to /src on the container (which is the default working directory).
  • Maps the Docker volume kitt-home to the kitt home directory on the container. This volume exists on the host and is created on the first run. This is useful to persist the preferences between sessions, but is not required.
• run_jupyter.sh: Another script provided for convenience, which mimics the previous one except:
  • Runs a jupyter server by default. So no TurboVNC server is launched. Therefore, only the 8888 port is mapped.
• build.sh: Script to build the image from the Dockerfile.
• Dockerfile: File used to build the image. This image is hosted on Docker Hub as yrahal/udacity-carnd.