东风汽车端到端自动驾驶数据集

业内领先的端到端公开数据集，聚焦车道线检测与3D目标检测两大核心场景，基于数据联盟标准《BEV目标检测标注规范》精心打造，为自动驾驶算法研发提供的高质量数据支撑。数据集收录10万组高精度标注数据、90万组自动标注数据，覆盖10类一级目标分类，包含丰富的动静态样本，所有 BEV 数据均通过同源采集与标注，确保数据一致性与模型泛化能力。数据涵盖晴天、雨天、雾天等全天气条件，白天、黄昏、夜间等多时间段，以及高架公路、城市道路等全类型路况，充分模拟真实驾驶环境。车道线标注精准反映道路结构，目标检测标注严格遵循行业规范，为车道保持、障碍物识别等关键任务提供可靠数据基础。

A leading end-to-end open dataset focusing on two core autonomous driving scenarios: lane detection and 3D object detection. It is meticulously developed based on the *BEV Object Detection Annotation Specification* from the data alliance standard, providing high-quality data support for autonomous driving algorithm research and development. The dataset includes 100,000 sets of high-precision manually annotated data and 900,000 sets of automatically annotated data, covering 10 primary target categories with abundant dynamic and static samples. All BEV data are collected and annotated from the same source, ensuring data consistency and model generalization capability. The data covers all weather conditions such as sunny, rainy, and foggy days, multiple time periods including daytime, dusk, and nighttime, as well as various road conditions like elevated highways and urban roads, comprehensively simulating real-world driving environments. The lane line annotations accurately reflect the road structure, and the object detection annotations strictly adhere to industry specifications, providing a reliable data foundation for key tasks such as lane keeping and obstacle recognition.

搜集汇总

背景与挑战

背景概述

该数据集是东风汽车提供的端到端自动驾驶公开数据集，专注于车道线检测和3D目标检测两大核心场景。它包含10万组高精度标注和90万组自动标注数据，覆盖10类目标分类，并涵盖多种天气、时间段和路况条件，以模拟真实驾驶环境。数据集基于行业标准构建，旨在为自动驾驶算法研发提供高质量数据支撑。