PointWire 和 PointVessel

本研究首次引入了两个针对三维可变形线性对象（DLOs）的点云数据集：PointWire 和 PointVessel。PointWire 基于40个真实扫描的汽车线束，通过半自动数据集生成器扩展至12000个样本，用于研究线束的拓扑结构和解缠。PointVessel 则源自136个血血管体积，转换为点云数据，用于医学影像领域中血管系统的研究。这两个数据集不仅促进了自动化线束制造的研究，还通过提供不同视角，推动了医学影像领域对血管系统的研究，特别是在缺乏数据的情况下，支持了制造和医学影像领域的迁移学习应用。

For the first time, this study introduces two point cloud datasets focused on three-dimensional (3D) deformable linear objects (DLOs): PointWire and PointVessel. PointWire is built upon 40 real-scanned automotive wire harnesses, and expanded to 12,000 samples via a semi-automatic dataset generator, designed for researching the topology and untangling of wire harnesses. PointVessel is derived from 136 blood vessel volumes, converted into point cloud data, and targeted at vascular system research in the medical imaging field. These two datasets not only advance research on automated wire harness manufacturing, but also promote vascular system studies in the medical imaging domain by providing diverse perspectives. Particularly in data-scarce scenarios, they support transfer learning applications in both manufacturing and medical imaging fields.