Topo-boundary

Topo-boundary是一个用于自主驾驶中拓扑道路边界检测的大型基准数据集，由香港科技大学创建。该数据集包含25,295张1000×1000大小的四通道航空图像，每张图像配有8个训练标签，用于不同的子任务。数据集的创建涉及从纽约市测绘数据库中提取原始地理数据，并将其转换为图像坐标系统，以便于研究。Topo-boundary数据集适用于基于分割和图形的道路边界检测解决方案，旨在通过提供大规模、高质量的数据集，推动自主驾驶领域的研究进展。

Topo-boundary is a large-scale benchmark dataset for topological road boundary detection in autonomous driving, developed by The Hong Kong University of Science and Technology. This dataset contains 25,295 four-channel aerial images with a resolution of 1000×1000, each paired with 8 training labels for different subtasks. The creation of the Topo-boundary dataset involves extracting raw geographic data from the New York City mapping database and converting it into the image coordinate system to facilitate relevant research. The Topo-boundary dataset is suitable for segmentation- and graph-based road boundary detection solutions, and aims to advance research progress in the autonomous driving field by providing a large-scale, high-quality benchmark dataset.