Today Tesla Motors has published another video demonstrating their Full Self-Driving features. This time it is more technical and more… familiar?
Is Tesla’s Autopilot rocket science? I’m sure it is. It took a lot of hours, days, months of Great Minds to build it. It invloves advanced Computer Vision, Deep Learning, Sensor Fusion and god knows what else.
But what I see in the middle right area (medium range vehicle camera) is so similar to the first project I’ve finished a couple of weeks ago at Udacity’s Self-Driving Car Engineer Nanodegree (https://www.udacity.com/drive).
It is way less robust to difficult road conditions than Tesla’s. It is sometimes not as accurate as it could be. But I was able to get this result after a week or so of learning. I cannot say it was easy. I rewrote the whole pipeline for drawing these lines 4 times from scratch. Every time I got results faster and my solution became cleaner. I read a whole bunch of articles and tutorials, watched numerous videos on YouTube on the topic.
But what helped me most was priceless communication with fellow Nanodegree students in Slack and at program’s forum. It was the main source of new ideas after my own ones were exhausted. And my wonderful mentor helped me to fine-tune my solution so that lines became much less noisy and jumpy than they were in the very beginning.
And the most exciting thing is I had no idea how computer vision works before I started the Nanodegree.
I encourage everyone interested in self-driving cars to try to solve this interesting problem. You don’t even have to sign up for the Nanodegree. Just clone this repo https://github.com/udacity/CarND-LaneLines-P1 and install required libraries.
But if you sign up you’ll get access to learning content of unbeliavable quality. Videos in courses I took at Udacity were always great. But this program is definitely new level.
Just look at this:
It can be as great and interesting as some of your favorite YouTuber’s content! And it is definitely more useful.
Of course, there are a lot of things in autonomus vehicles field beyond lane lines detection. Lots of them are covered by the amazing 9-months curriculum: https://medium.com/self-driving-cars/term-1-in-depth-on-udacitys-self-driving-car-curriculum-ffcf46af0c08#.3t1i35gbc
I would like to finish this post by mentioning hiring partners of this Nanodegree. As you can notice there is no Tesla (c’mon man, aren’t you satisfied with Mercedes-Benz and McLaren?). But based on what I saw in Tesla’s recent Demo (the video in the beginning of the post) and what I saw at Udacity classroom (it’s where all learning content is, and the last video is an example of that) I believe Nanodegree graduates are going to be eligible for positions in Tesla, Mercedes-Benz, Otto or any other startup or company of their choice.
