FloorPlanCAD

FloorPlanCAD是一个大规模的实际CAD绘图数据集，包含超过15,000个从住宅到商业建筑的平面图。该数据集中的CAD绘图均以矢量图形表示，支持35个对象类别的线粒度标注。数据集从公共存储库中收集，涵盖了多种建筑类型，如住宅塔楼、学校、医院、购物中心和办公楼。创建过程包括从原始的“.dwg”文件解析到开放标准“.svg”格式，并将其分割成规则的小块。该数据集主要用于解决全景符号定位问题，通过结合图卷积网络(GCNs)和卷积神经网络(CNNs)来实现高效的3D建模方法。

FloorPlanCAD is a large-scale real-world CAD drawing dataset containing over 15,000 floor plans ranging from residential to commercial buildings. All CAD drawings in this dataset are represented in vector graphics, with line-level annotations covering 35 object categories. The dataset is collected from public repositories, covering a diverse range of building types such as residential towers, schools, hospitals, shopping malls and office buildings. The creation process includes parsing original ".dwg" files into the open-standard ".svg" format, followed by splitting them into regular small chunks. This dataset is primarily developed to address the panoramic symbol localization task, enabling efficient 3D modeling methods by combining Graph Convolutional Networks (GCNs) and Convolutional Neural Networks (CNNs).

搜集汇总

背景与挑战

背景概述

FloorPlanCAD是一个大规模真实世界CAD图纸数据集，包含超过15,000个建筑平面图，涵盖住宅和商业类型，以SVG格式提供并带有线粒度标注。该数据集专门设计用于支持'全景符号识别'任务，旨在识别可计数实例和不可计数材料的语义，适用于计算机视觉和图形算法开发。