CurveLanes

该数据集名为CurveLanes，是一个处理包括分叉和密集线条等复杂拓扑结构的新型车道检测基准数据集。该数据集被划分为训练集（10万张）、验证集（2万张）和测试集（3万张），每张图片的分辨率为2650×1440像素。作为世界上最大的车道检测数据集，它包含了15万张车道图片。该数据集的任务是进行车道检测。

The dataset named CurveLanes is a novel lane detection benchmark dataset that addresses complex topological structures including bifurcations and dense lane lines. It is divided into three subsets: a training set with 100,000 images, a validation set with 20,000 images, and a test set with 30,000 images. Each image has a resolution of 2650×1440 pixels. As the largest lane detection dataset worldwide, it contains a total of 150,000 lane-related images. The core task of this dataset is lane detection.

搜集汇总

背景与挑战

背景概述

CurveLanes是一个用于车道检测的基准数据集，包含150K张图像，专注于曲线和多车道等复杂场景，采集自中国多个城市的真实环境。数据集分为训练集、验证集和测试集，图像分辨率高，标注精细，旨在为车道检测研究提供更具挑战性的基准。