The Oxford Road Boundaries Dataset

The Oxford Road Boundaries Dataset是由牛津机器人研究所创建的一个专注于道路边界检测的数据集。该数据集包含62605个标注样本，其中47639个样本经过精心筛选。数据集内容丰富，涵盖了多种道路场景，如直路、停车区、交叉口等，每个样本均包含左右镜头的原始和分类掩码。创建过程中，利用视觉定位工具将标注从已标注数据集投射到其他时间、天气条件下的行驶数据中，以增加训练样本的数量。该数据集主要应用于自动驾驶领域，旨在提高车辆对道路边界的检测和理解能力，确保行车安全。

The Oxford Road Boundaries Dataset was developed by the Oxford Robotics Institute, focusing on road boundary detection tasks. This dataset comprises 62,605 annotated samples, among which 47,639 samples were carefully curated. It covers a wide range of road scenarios including straight roads, parking areas, intersections and more. Each sample contains raw images and categorized masks captured by both left and right cameras. During the dataset construction, visual localization tools were employed to project annotations from the pre-annotated dataset into driving data collected under different temporal and weather conditions, thereby augmenting the number of training samples. This dataset is primarily utilized in the autonomous driving domain, with the goal of enhancing vehicles' capability to detect and comprehend road boundaries and ensuring driving safety.