The video demonstrates semantic segmentation for road detection, obstacles detection, and lane markers and road delimiters detection jointly via our DGN-I and LDG Networks.
LDG detects lane markers if they exist, and automatically generates them where they ideally should have been, if such lane markers or road delimiters are non-existent, or faded or not visible in the images. All of this happens in real-time at only 12.5 GFlops per image.