V2V4Real
收藏资源简介:
V2V4Real是由加州大学洛杉矶分校开发的第一个大规模真实世界多模态数据集,专注于车辆间协同感知。该数据集覆盖了410公里的驾驶区域,包含20,000个LiDAR帧、40,000个RGB帧和240,000个标注的3D边界框,适用于5类车辆。数据集的创建过程涉及两辆配备多模态传感器的车辆在多样化场景中同时行驶,并通过专业标注团队进行精确标注。V2V4Real数据集的应用领域主要集中在解决自动驾驶中的长距离感知和遮挡问题,通过车辆间通信和数据共享,显著提升感知系统的性能。
V2V4Real is the first large-scale real-world multimodal dataset developed by the University of California, Los Angeles (UCLA) for vehicle-to-vehicle cooperative perception. This dataset covers a driving area of 410 kilometers, containing 20,000 LiDAR frames, 40,000 RGB frames, and 240,000 annotated 3D bounding boxes for 5 vehicle categories. It was created by having two vehicles equipped with multimodal sensors driving simultaneously in diverse scenarios, with precise annotation conducted by professional annotation teams. The primary application scenarios of V2V4Real focus on addressing long-distance perception and occlusion issues in autonomous driving, and it can significantly improve the performance of perception systems through vehicle-to-vehicle communication and data sharing.




