DeepScenario Open 3D Dataset (DSC3D)
收藏资源简介:
DeepScenario Open 3D数据集(DSC3D)是一个高质量、无遮挡的6自由度边界框轨迹数据集,通过创新的单目相机无人机跟踪流程获取。该数据集包含超过175,000个轨迹,涉及14种交通参与者类型,在多样性和规模上显著超过了现有数据集,包括复杂的车辆-行人交互和全面的停车操作。数据集在欧洲和美国的五个不同地点捕获,包括停车场、繁忙的内城、陡峭的城市交叉口、联邦公路和郊区交叉口。DSC3D数据集旨在通过提供详细的环境3D表示来增强自动驾驶系统,这可能有助于提高障碍物交互和安全。该数据集在多个应用程序中展示了其效用,包括运动预测、运动规划、场景挖掘和生成反应交通代理。交互式在线可视化平台和数据集的完整数据都是公开可用的,以促进运动预测、行为建模和安全验证的研究。
DeepScenario Open 3D Dataset (DSC3D) is a high-quality, occlusion-free 6-degree-of-freedom (6DoF) bounding box trajectory dataset acquired through an innovative monocular camera-based drone tracking pipeline. This dataset contains over 175,000 trajectories and encompasses 14 categories of traffic participants, substantially outperforming existing datasets in both diversity and scale, covering complex vehicle-pedestrian interactions and comprehensive parking maneuvers. The dataset was collected at five distinct sites across Europe and the United States, including parking lots, dense urban downtowns, steep urban intersections, federal highways, and suburban intersections. Designed to advance autonomous driving systems by providing detailed 3D environmental representations, DSC3D can facilitate improved obstacle interaction and driving safety. The utility of this dataset has been validated across multiple applications, including motion prediction, motion planning, scene mining, and the generation of reactive traffic agents. An interactive online visualization platform and the complete dataset are publicly available to foster research on motion prediction, behavior modeling, and safety validation.

- 1Highly Accurate and Diverse Traffic Data: The DeepScenario Open 3D DatasetDeepScenario, TU Munich, Munich Center for Machine Learning · 2025年



