GauU-Scene V2
收藏资源简介:
GauU-Scene V2是由香港中文大学(深圳)创建的大型多模态场景重建基准数据集,覆盖面积超过6.5平方公里,包含全面的RGB数据和LiDAR地面实况。该数据集通过高精度LiDAR和无人机搭载的高分辨率相机收集,旨在解决大规模户外场景的3D重建问题。数据集的创建过程中,提出了一种LiDAR和图像对齐方法,以解决无人机数据集中的坐标系统差异问题。GauU-Scene V2的应用领域包括城市规划、虚拟现实和增强现实,通过评估不同视角下的图像质量,揭示了当前基于图像的度量标准在几何重建中的局限性。
GauU-Scene V2 is a large-scale multimodal scene reconstruction benchmark dataset developed by The Chinese University of Hong Kong, Shenzhen. It covers an area of over 6.5 square kilometers and contains comprehensive RGB data and LiDAR ground truth. Collected via high-precision LiDAR and high-resolution cameras mounted on unmanned aerial vehicles (UAVs), this dataset is designed to address 3D reconstruction challenges in large-scale outdoor scenes. During the dataset's development, a novel LiDAR-image alignment method was proposed to resolve coordinate system discrepancies inherent in UAV-based datasets. Application scenarios of GauU-Scene V2 include urban planning, virtual reality (VR) and augmented reality (AR). By evaluating image quality across different viewpoints, this benchmark reveals the limitations of current image-based metrics in geometric reconstruction tasks.




