Our Lanes

Our Lanes数据集由FiveAI Ltd.创建，专注于城市环境中的车道实例分割。该数据集包含23,980张图像，每张图像平均标注时间为5秒，使用低成本的行车记录仪设备进行数据采集。数据集的创建过程采用半自动化标注方法，通过利用车辆行驶轨迹在3D空间中估计道路平面，并将标注从该平面投影到序列中的所有图像上，有效减少了手动标注的时间和成本。该数据集主要应用于自动驾驶技术中，旨在解决车道检测和分割的问题，特别是在复杂的城市交通场景中。

Our Lanes dataset was created by FiveAI Ltd., focusing on lane instance segmentation in urban environments. This dataset contains 23,980 images, with an average annotation time of 5 seconds per image, and was collected using low-cost dashcam equipment. The dataset was developed via a semi-automated annotation workflow: it estimates the road plane in 3D space using vehicle driving trajectories, then projects the annotations from this plane onto all images in the sequence, effectively reducing manual annotation time and costs. This dataset is primarily applied in autonomous driving technologies, aiming to address lane detection and segmentation challenges, particularly in complex urban traffic scenarios.