it's part of self-driving stack still under development, visual odometry extract depth map, odometry, enhance detection, localization certainty, and lane keeping with aid of semantic segmentation.
extract vertical lines, keep those detected as lanes, filter the current lane area, estimate distance to object from stereo.
extract the disparity map
