Simultaneous Road Edge and Road Surface Markings Detection Using Convolutional Neural Networks.

Accurate road surface markings and road edges detection is a crucial task for operating self-driving cars and for advanced driver assistance systems deployment (e.g. lane detection) in general. This research proposes an original neural network based method that combines structural components of autoencoders, residual neural networks and densely connected neural networks. The resulting neural network is able to concurrently detect and segment accurate road edges and road surface markings from RGB images of road surfaces.