My 2018 Path to Learning.
Term 1 Cohort of my Self Driving Car Nanodegree classroom starts on the 19th, meanwhile I have started working on things in my Garage that I want to share:
2 Weeks in:
As Promised, I would add posts to the series regularly
Here is some news:
I have decided to make a Mini Self Driving Car to apply what I will learn in my Self Driving Car Nanodegree Program:
- The Setup: The Model 1 (Name TBD-Please help me pick a name in the comments below!) will be Raspberry Pi 3 based + RC Car setup.
- Reason: Raspberry Pi 3 although has less horsepower (pun intended) to run all the Image Processing and detection tasks is a good place to test my skills as a primer and certainly later, I will apply the same to a Nvidia Jetson TX-2 (Or 1). The secondary reason being, I don’t own a Jetson and I will apply for Student aid options later.
1. I’ve found a RC Toy Car (from my childhood), that will function as the main chassis.
2. Setup RPI3 board: I’ve installed Ubuntu Mate on the Raspberry Pi Board, configured ROS and OpenCV on it.
3. Setup the PiBoard for Remote Logins and running scripts on boot: which is essential, since in the field my Model 1 car should be able to boot up and drive right away!
I have also taken the courage to set 5 bold targets before my Term 1 starts (Roughly a week from now):
- Complete 2 Projects before the Term’s opening:
I will attempt to make a crude solution to the Lane Detection & Sign Classification Projects, before the Nanodegree opens! (The Project Repositories are hosted on GitHub and I will attempt to solve it using knowledge and hints provided via them). Later, I’ll share a proper review once the Projects ‘Meet specifications’-which is Udacity’s codeword for saying Project Complete!
- Deploy Lane Detection on Model 1 Mini SDC (Guys, please help me pick a name, this doesn’t sound cool at all). Since the Lane Detection pipeline is OpenCV based, I’ll test on the RPI board using the Pi Camera.
- Create a Simulation of a Self Driving Car using ROS:
I will not Udacity’s Simulator, I will take a shot at Gazebo and use Open Source Models of Cars to make a simulation. The goal is to make a model where I can play around with algorithms (not to make a fully functioning prototype inside a simulation)
- Learn Keras basics:
I have worked on Tensorflow Projects in my Deep Learning Nanodegree and in my Internship Projects, PyTorch in the FastAI International Fellowship. Next up is Keras which will be used in my Term 1.
By no means am I a master of these Libraries, but I’m keen on learning and implementing Models using them.
- Complete Lex Fridman’s Self Driving Course at MIT’s 2018 version.
Long Term functionality Goals for My Mini Self Driving Car:
- Path (lane) following
- Crash Avoidance
- Object Detection
- Symbol Detection
- Bonus points: A drift mode!
Please note that these aren’t the Pre-requistes, rather are my own personal ventures to my Self Driving Year.
I will share detailed posts of the targets that I have completed.
Stay Tuned for updates! I will keep sharing my story, if I make it to the finish line with my Goals or hit the dust!
These series will also be a public accountability of how I have been working on my Journey and hence the title for the series ‘A Self Driving Year’