[billyzs]

Not to put down the OP's work (I think it's a great project), but I'm just wondering what advantages might an ML approach have over "traditional" CV algorithms. In a really well controlled environment lanes will be easy to detect, and computing the difference between the current heading and lane direction should be doable; maybe if we're talking about complex outdoor environments and poor sensors then ML would have an advantage? Or if we're teaching the robot what the concept of a lane is?

I think back to the days when I basically implemented lane following with an array of photo resistors, an Arduino, a shitty robot made from Vex parts and some c code. The problem is much simpler than the one presented in this article, but then the computational resource used was order of magnitudes less. At what point then, do you decide that "OK I think the complexity and nature of the problem warrants the use of ML" or "Hmmm I think neural network is an overkill here"?

[wroscoe]

Traditional CV approaches are much easier to debug as well. I chose the ML approach with the assumption that it would be easier to build a robust autopilot that would work in many lighting conditions. Actually my short term goal is to get the car to drive around my block on the sidewalk (no lines). From my experience CV approaches have many parameters that need to be tuned specifically for each environment. While ML approaches also have parameters that need tuning they stay constant between environments.

[billyzs]

I see, that makes sense. It'd be indeed worth it if we can apply a model trained on controlled environment to a more challenging one with little to no modification. Good luck with the project and keep us updated!

[option_greek]

Because ML approaches can adapt to different environments like a forest trail. While this can probably be achieved with OpenCV, this just feels natural: https://www.youtube.com/watch?v=umRdt3zGgpU

[platz]

My first thought was something that used several PID mechanisms.