Looking at the code, the good stuff is in the binary blob. The binary blob returns polygons for the edges of the road, with a probability for each. Path planning, in "pathplanner.py", consists of a probability-weighted average to find the center of the lane, using an assumption of 3.6m lane width. There's no map-building at all. It's purely reactive.

Speed control uses data from a radar and from the vision system. The path info is used to decide which targets the radar should consider important. The big thing seems to be identifying the "car ahead", if any, and maintaining an appropriate distance.

There's a hard-coded 6 minute timeout to make the driver take control once in a while.

On a good day on a well-marked freeway, this might work. It would be interesting to modify this to process pre-recorded dashcam data and see how well it tracks the road.

Based on a really cursory inspection of the binary and symbols, the binary seems to be an OpenCV project which uses some common image processing algorithms (Sobel edge detection, Hough to find straight lines).

There doesn't seem to be a large amount of embedded data indicative of a trained model or neural network.

So, it seems that the vision algorithms probably mirror those in the common literature and represent a reimplementation or iterative improvement on existing lane-finding systems.

More reversing would probably yield greater insight :)

visiond appears to be linked against libgsl.so. To my understanding, distributing a binary blob using the GNU Scientific Library is against the license of this library. That written there does not seem to be any symbol used from libgsl.so.