houselayout3d/HouseLayout3D

HouseLayout3D是一个面向大规模多楼层建筑的三维布局估计的挑战性基准数据集。它基于Matterport3D的实际建筑扫描构建，包含多达五层楼和每个建筑四十个房间的详细建筑元素标注。数据集旨在支持场景理解、室内映射和需要向量化、无对象室内空间表示的机器人应用研究。

HouseLayout3D is a challenging benchmark dataset for 3D layout estimation in large-scale, multi-floor buildings. It is built upon real-world building scans from Matterport3D and provides detailed annotations of structural elements across up to five floors and forty rooms per building. The dataset is designed to support research in scene understanding, indoor mapping, and robotics applications that require vectorized, object-free representations of indoor spaces.