Software for Term 1 of the Udacity Self-Driving Car Engineer Nanodegree. Python 3 is used for entirety of term 1.

There are two ways to get up and running:

1. Anaconda Environment
2. Docker

Install miniconda on your machine.

Next, setup the CarND term 1 environment.

To install:

git clone https://github.com/udacity/CarND-Term1-Starter-Kit.git
cd CarND-Term1-Starter-Kit
conda env create -f environment.yml

Verify that the carnd-term1 environment was created in your environments:

conda info --envs

If the carnd-term1 environment is not listed, make sure that the environment.yml file exists in your directory "..\CarND-Term1-Starter-Kit". If it does not exist there, the previous conda command (conda env create -f environment.yml) won't create the carnd-term1 environment. You can manually add the environment.yml file by downloading it from the git repository (https://github.com/udacity/CarND-Term1-Starter-Kit/) using your browser and adding the file to the directory "..\CarND-Term1-Starter-Kit". Perform the conda command "conda env create -f environment.yml" again if necessary.

To use:

# Enter the environment, should be called before you plan to work on a project (unless already active)
source activate carnd-term1
# Exit the environment
source deactivate

To cleanup downloaded libraries (remove tarballs, zip files, etc):

conda clean -tp

To uninstall the environment:

conda env remove -n carnd-term1

The current setup only installs the CPU version of TensorFlow. If you wish to use the GPU version follow the instructions here.

Using Docker to run your code consists of the following:

1. Install Docker on your computer
2. Pull the precompiled Docker image from Docker Hub
3. Run the image as a new container

You may also wish to run a [python module][doc/py_mod.md] or [ipython][doc/ipython.md].

Instructions for installation very by operating system and version.

OS Specific instructions can be found below:

• Docker for Linux
  • Linux
• Docker for Mac
  • MacOS >= 10.10.3 (Yosemite)
• Docker Toolbox for Max
  • MacOS >= 10.8 (Mountain Lion)
• Docker for Windows
  • Windows 10 Pro, Enterprise, or Education
• Docker Toolbox for Windows
  • Windows 7, 8, 8.1, or 10 Home

Recommended Shell:

A precompiled image with all dependencies required for the first term is available on Docker Hub.

Once you have docker working, pull the image using the following command:

docker pull udacity/carnd-term1-starter-kit

In your shell, navigate to the directory of a project, e.g.

$ cd ~/src/CarND-LaneLines-P1

From within this directory, you are going to run a Jupyter server. In order to do this you must attach to the correct port and share a local volume.

The easiest way to share a local volume is via the pwd command, a shell command that prints the working directory. This command will be used differently based on your shell.

If you're using Windows PowerShell:

docker run -it --rm -p 8888:8888 -v ${pwd}:/src udacity/carnd-term1-starter-kit

If you're using bash or Docker Quickstart Terminal:

docker run -it --rm -p 8888:8888 -v `pwd`:/src udacity/carnd-term1-starter-kit

Let's break this down.

docker run is the command a startup and run a Docker container.

-it forces the container to run in the foreground (interactive mode) and provides an I/O to the container.

--rm removes the container once it stops running. It prevents the buildup of stale containers once you stop them from running.

-p 8888:8888 maps port 8888 on our local machine to port 8888 in the Docker container, this allows us to access port 8888 in the container by visiting localhost:8888.

-v ${pwd}:/src mounts the pwd (present working directory) to the /src directory in the container. Basically, this let's us access files from our local machine on the docker container.

udacity/carnd-term1-starer-kit is the name of the container to run.

To learn more about Docker visit the docs.

The current image does not support GPU use. An image with GPU support is in the works although this would only work with a Linux base OS.