CRUW
收藏资源简介:
CRUW数据集是由华盛顿大学和浙江大学等机构联合创建的大规模多模态数据集,专注于雷达目标检测任务。该数据集包含约40万条同步的摄像头雷达帧,覆盖多种驾驶场景,如停车场、校园道路、城市街道和高速公路。数据集通过精确的坐标对齐和系统化的标注方法,旨在从雷达的射频图像中纯3D地分类和定位物体。此外,CRUW还引入了一套评估指标,包括对象位置相似度(OLS)和检测质量F1分数(DQF1),以全面评估雷达目标检测的性能。数据集的应用领域主要集中在自动驾驶和辅助驾驶系统中,特别是在恶劣天气和光照条件下的物体感知和分类。
The CRUW dataset is a large-scale multimodal dataset jointly developed by the University of Washington, Zhejiang University, and other institutions, focusing on radar object detection tasks. It contains approximately 400,000 pairs of synchronized camera and radar frames, covering a wide range of driving scenarios including parking lots, campus roads, urban streets, and highways. Through precise coordinate alignment and systematic annotation methods, this dataset aims to conduct purely 3D classification and localization of objects from radar radio frequency (RF) images. In addition, CRUW also introduces a standardized set of evaluation metrics, including Object Location Similarity (OLS) and Detection Quality F1 Score (DQF1), to comprehensively evaluate the performance of radar object detection systems. The primary application domains of this dataset are autonomous driving and advanced driver-assistance systems (ADAS), particularly for object perception and classification under adverse weather and lighting conditions.

- 1Rethinking of Radar's Role: A Camera-Radar Dataset and Systematic Annotator via Coordinate Alignment华盛顿大学 · 2021年



