Generating All the Roads to Rome: Road Layout Randomization for Improved Road Marking Segmentation
This addresses the need for scalable and efficient training data generation in automated driving, particularly for rare road marking classes, though it is an incremental improvement over existing domain randomization methods.
The paper tackles the problem of costly and time-consuming manual labeling for road marking segmentation by automatically generating synthetic training pairs through domain randomization of road layouts. The result is a more than 12 percentage point improvement in mIoU for rare road marking classes in real-world urban environments, while maintaining performance for other classes.
Road markings provide guidance to traffic participants and enforce safe driving behaviour, understanding their semantic meaning is therefore paramount in (automated) driving. However, producing the vast quantities of road marking labels required for training state-of-the-art deep networks is costly, time-consuming, and simply infeasible for every domain and condition. In addition, training data retrieved from virtual worlds often lack the richness and complexity of the real world and consequently cannot be used directly. In this paper, we provide an alternative approach in which new road marking training pairs are automatically generated. To this end, we apply principles of domain randomization to the road layout and synthesize new images from altered semantic labels. We demonstrate that training on these synthetic pairs improves mIoU of the segmentation of rare road marking classes during real-world deployment in complex urban environments by more than 12 percentage points, while performance for other classes is retained. This framework can easily be scaled to all domains and conditions to generate large-scale road marking datasets, while avoiding manual labelling effort.