PAD-Highway
收藏资源简介:
PAD-Highway 数据集是一个用于个性化自动驾驶行为分析的闭循环基准,基于 Highway-Env 模拟器构建。该数据集包含 250 小时的高质量视频,并带有详细标注,旨在促进个性化自动驾驶的研究。数据集由规则驱动和人类驱动两部分组成,共收集了 235 小时的规则驱动数据和 25 小时的人类驾驶数据,包含 32,000 个视频片段。数据集提供了自动驾驶车辆的环境信息、动作、速度和坐标等参数,可用于训练和评估个性化自动驾驶模型。
The PAD-Highway dataset is a closed-loop benchmark for personalized autonomous driving behavior analysis, built on the Highway-Env simulator. This dataset contains 250 hours of high-quality videos with detailed annotations, aiming to facilitate research on personalized autonomous driving. The dataset consists of two parts: rule-driven and human-driven. A total of 235 hours of rule-driven data and 25 hours of human driving data have been collected, including 32,000 video clips. The dataset provides parameters including environmental information, actions, speed and coordinates of autonomous vehicles, which can be utilized to train and evaluate personalized autonomous driving models.

- 1PADriver: Towards Personalized Autonomous Driving北京理工大学、Megvii Technology、早稻田大学 · 2025年



