UniSPC, RealSPC
收藏资源简介:
UniSPC和RealSPC是由清华大学和剑桥大学的研究团队基于BIM技术生成的合成点云数据集。UniSPC数据集采用一致的BIM颜色,而RealSPC数据集则使用真实颜色。这些数据集通过模拟扫描过程生成,旨在弥补真实点云数据的不足,并为深度学习模型的训练提供高质量数据。数据集的内容包括建筑环境的点云信息,数据量未明确提及,但通过与S3DIS数据集的结合进行了实验验证。数据集的生成过程涉及几何生成、语义标注和颜色分配三个步骤,特别关注了颜色对模型性能的影响。这些数据集主要用于建筑和施工领域的语义分割任务,旨在提升计算机对复杂建筑环境的理解能力。
UniSPC and RealSPC are synthetic point cloud datasets generated by research teams from Tsinghua University and the University of Cambridge based on BIM technology. The UniSPC dataset adopts consistent BIM-compliant colors, while the RealSPC dataset uses real-world colors. These datasets are generated by simulating the scanning process, aiming to compensate for the scarcity of real-world point cloud data and provide high-quality training data for deep learning models. The datasets cover point cloud information of built environments, and their specific scale is not explicitly specified, but they have been experimentally validated through integration with the S3DIS dataset. The generation workflow of the datasets involves three core steps: geometry generation, semantic annotation, and color assignment, with particular attention paid to the impact of color configurations on model performance. These datasets are primarily used for semantic segmentation tasks in the architecture, engineering and construction (AEC) industry, with the goal of enhancing computers' capability to comprehend complex built environments.

- 1Impact of color and mixing proportion of synthetic point clouds on semantic segmentation清华大学,剑桥大学 · 2024年



