Detect Lane Lines on the Road

This is my first project in the Udacity Nanodegree program Self-Driving Car Engineer, and also my first project using Git and GitHub together!

The original readme.md is a manual for students; I change it to a documentation of how I got the set-ups going. If any student would have difficulties gettig started, and happend to find my readme.md, I hope she or he could find my experience useful.

Note: My thoughts on the implementation of the project are written in writeup_template.md.

Prio to the project

Prio to this project, I took two Udacity free courses Version Control with Git and Writing readmes, which were very helpful to me to get to know Git. However, they did not teach students how Git and GitHub work together. I used the search engine to teach myself.

For this project, I did the following set-ups:

• Using git clone <repository URL> to get the repo from Udacity GitHub
• On my GitHub, creating a new repo with no initializing, and copying the URL
• Using git remote set-url origin <repository URL> to change the origin to my GitHub repo
• Using git push -u origin master to upload the repo to my GitHub

Note: During the course I used Atom, which I find perfect to write readme.md, because I can edit the Markdown text while seeing the preview!

Starter-Kit

Udacity provides a starter-kit to students for a unified environment. When I was following the instructions, I got a problem creating the environment using conda env create -f environment.yml. It took a dozen minutes only ending up with no environment (which you can see by trying to activate the environment source activate carnd-term1). Thanks to a question in the Udacity Knowledge, I could solve this by putting all entries of "dependencies" under "-pip:" in the environment.yml in the starter-kit folder (see below; additionally, I removed all versions in this .yml as suggested by the answers).

Jupyter notebook

Then I tried to run jupyter notebook, but the terminal said that this command did not exist. I thought Jupyter notebook is automatically installed within Miniconda; it turned out that I needed to install it myself. So I ran conda install -c conda-forge notebook to get the notebook. Then it worked!

Python modules

So I opened the code file and tried to import Python modules. Again, an error message... No modules were found. OK, let me get the modules in Miniconda:

• conda install matplotlib which also got numpy for me.
• conda install -c menpo opencv3, again, error message. My Python version is too high for opencv3.
• Wait a minute, opencv's latest version is 4.2.0, why am I installing opencv3? On the conda-forge website I found the command to install 4.2.0: conda install -c conda-forge opencv
• conda install -c conda-forge moviepy on the conda-forge moviepy 1.0.1 site

Now I can import all modules without an error message.

Image functions

The function image reading returned an error saying: "ValueError: Only know how to handle extensions: ['png']; with Pillow installed matplotlib can handle more images". Yep, it cannot handle the jpg files. Hey, since I have opencv, why not just use cv2.imread? It worked!

The function image showing returned an error saying: "TypeError: Image data of dtype object cannot be converted to float". Again, I have opencv, so I tried cv2.imshow. I had to use cv2.waitKey(0) and cv2.destroyAllWindows() together to get the image, otherwise the window just freezed and I hat to restart Jupyter notebook.

The function image writing only worked, after the folder was created manually.

Summary

It was not easy at all to got all set-ups correct. As a newcomer I encounted the above problems and had to do a lot of search to find solutions. But I am very glad that I have overcome all kinds of trouble!

Note: The Udacity tutors have made 92 commits. The commits from 93 are from me :)