Self-Driving Car Nanodegree students in their own words
I’ve been extremely impressed by our students in the Self-Driving Car Nanodegree so far. They are not only creating great projects that go above and beyond what we ask of them, but also communicating their ideas through extremely well written blog posts. Today, I want to share some of the posts to give a sense of the amazing work students in the Self-Driving Car Nanodegree are capable of. Thank you for your posts Param Aggarwal, Thomas Antony, Mehdi Sqalli, Jessica Yung, Aaron Smith, and others!
Project 1 — Finding Lane Lines
Author: Param Aggarwal
Param not only created a successful lane detection pipeline, but also was able to get it to work on roads in the notoriously difficult driving conditions of India! You can read through his post for a wonderful, step by step guide on how he did it and what he plans to do next.
Author: Aaron Smith
Blog Post: http://flippidybits.com/project-1-lane-finding/
Aaron Smith, one of our most active community members, created a similarly excellent guide to his solution for finding lane lines. Note how he succeeds on lane lines of different colors and links out to further resources for you to learn more!
Author: Jessica Yung
Blog Post: http://www.jessicayung.com/bugger-detecting-lane-lines/
Jessica is one of the most organized students I’ve seen! All her code for the Nanodegree is in a single, neatly organized github. What I particularly love about this post is how she works through the bugs she sees and helps others understand why those bugs occur and how one can address them.
Project 2 — Identifying Traffic Signs with Deep Learning
Author: Mehdi Sqalli
Blog Post: https://hackernoon.com/traffic-signs-classification-with-deep-learning-b0cb03e23efb#.fvnw67br4
Mehdi is not only doing the Self-Driving Car Nanodegree, but also running a YC startup, SpotAngels, which builds on-street parking maps for connected vehicles! In this post, Mehdi walks through both the intuition and engineering for using Deep Learning to classify traffic signs. We share Mehdi’s optimism for Deep Learning when he concludes that “Deep learning is really impressive. I can’t wait to see how much I can still learn and progress in this domain.”
Project 3 — Behavioral Cloning with Deep Learning
Author: Thomas Anthony
Thomas’ post was so impressive that Sebastian Thrun, founder of Udacity and the winner of the DARPA Grand Challenge, shared it with the world!
This project was by no means easy and Thomas persevered by chatting with his peers, using Udacity’s support resources, and just working really really hard. I was extremely impressed by how Thomas worked with another student, John Chen, to come up with a custom trainer for collecting data. This is well above and beyond what we did to solve the project! Thomas’ project is awesome. He shows how without programming any kind of driving rules, you can use deep learning to teach a car to drive itself around a track. What’s more impressive, Thomas’ model is able to drive perfectly on both tracks we presented in the simulator showing that he’s indeed created a model that generalizes.
