June 2020
tl;dr: Slope compensated lane line detection.
Nothing too impressive about this approach. The approach is not even end to end differentiable and uses a nonlinear optimizer for solution. This is not quite transferrable.
It only targets to solve 90% of the problem ("parallel polynomials") and still does not solve split or merge issues. --> See Semilocal 3D LaneNet for a method to solve more complex topologies.
- A novel loss that involves entropy and histogram. The main idea is that in BEV space the lane line points collapsed to the x dimension should have multiple equally spaced peaks. But this loss is not differentiable.
- Approximate a road slope. This is essentially the pitch estimation of the road in LaneNet.
- Summary of technical details
- Questions and notes on how to improve/revise the current work