SimBEV
收藏资源简介:
SimBEV数据集是由弗吉尼亚联邦大学的研究人员创建的一个大规模合成多任务多传感器驾驶数据集。该数据集通过SimBEV工具生成,能够支持多种感知任务,包括鸟瞰图(BEV)分割、3D对象检测等。数据集包含了多样化的驾驶场景,通过随机化多种仿真参数,生成具有统计多样性的场景。SimBEV工具支持多种传感器,包括不同类型的摄像头、雷达、GNSS和IMU等,并提供了3D对象边界框和BEV地面真实标签。该数据集旨在为自动驾驶领域的研究提供支持,助力开发更为先进和稳健的感知系统。
The SimBEV dataset is a large-scale synthetic multi-task and multi-sensor driving dataset created by researchers from Virginia Commonwealth University. It is generated using the SimBEV tool, and supports a variety of perception tasks including bird's-eye view (BEV) segmentation, 3D object detection, and more. The dataset encompasses diverse driving scenarios, which are produced by randomizing multiple simulation parameters to generate statistically varied scenes. The SimBEV tool supports various sensor modalities including different types of cameras, radar, GNSS, and IMU, and provides 3D object bounding boxes and BEV ground-truth labels. This dataset is intended to support research in the autonomous driving field and aid the development of more advanced and robust perception systems.




