Road-boundary detection is important for autonomous driving. For example, it can be used to constrain vehicles running on road areas, which ensures driving safety. Compared with on-line road-boundary detection using on-vehicle cameras/Lidars, off-line detection using aerial images could alleviate the severe occlusion issue. Moreover, the off-line detection results can be directly used to annotate high-definition (HD) maps. In recent years, deep-learning technologies have been used in off-line detection. But there is still lacking a publicly available dataset for this task, which hinders the research progress in this area. So in this paper, we propose a new benchmark dataset, named Topo-boundary, for off-line topological road-boundary detection. The dataset contains 21,556 1000*1000-sized 4-channel aerial images. Each image is provided with 8 training labels for different sub-tasks. We also design a new entropy-based metric for connectivity evaluation, which could better handle noises or outliers. We implement and evaluate 3 segmentation-based baselines and 5 graph-based baselines using the dataset. We also propose a new imitation-learning-based baseline which is enhanced from our previous work. The superiority of our enhancement is demonstrated from the comparison.

Please check the new project page.