ParallelEye
收藏资源简介:
ParallelEye数据集是由中国科学院自动化研究所创建的大规模人工交通场景数据集,旨在支持交通视觉研究。该数据集通过使用街道地图数据构建北京中关村地区的3D场景模型,并利用计算机图形学、虚拟现实和规则建模技术合成了大规模、真实的虚拟城市交通场景。数据集包含31,000帧图像,每帧图像都自动标注了包括语义/实例分割、物体边界框、物体跟踪、光流和深度在内的精确地面实况标签。ParallelEye数据集的应用领域主要集中在智能交通系统,特别是汽车导航和视觉监控,以解决现有数据集在规模和多样性上的不足。
The ParallelEye dataset is a large-scale artificial traffic scene dataset developed by the Institute of Automation, Chinese Academy of Sciences, aiming to support traffic vision research. It constructs a 3D scene model of the Zhongguancun area in Beijing using street map data, and synthesizes large-scale, realistic virtual urban traffic scenes through computer graphics, virtual reality and rule-based modeling technologies. The dataset contains 31,000 image frames, each of which is automatically annotated with precise ground-truth labels including semantic/instance segmentation, object bounding boxes, object tracking, optical flow and depth. The application fields of the ParallelEye dataset mainly focus on intelligent transportation systems, particularly vehicle navigation and visual surveillance, to address the shortcomings of existing datasets in terms of scale and diversity.




