Combining rotation-invariance images and neural networks for road scene understanding | IEEE Conference Publication | IEEE Xplore