Scan2CAD

该数据集精确地将ScanNet中的对象实例与ShapeNet中的CAD模型对齐，为多个真实扫描场景提供了准确的刚体变换标注。每个扫描场景包含2至5个相同的CAD实例，使得该数据集能够对点云配准方法进行全面的评估。数据集规模包括：1523个场景用于训练，326个场景用于验证，另外326个场景用于测试。该数据集的任务是多点云实例配准。

This dataset accurately aligns object instances from ScanNet with CAD models from ShapeNet, providing precise rigid transformation annotations for multiple real-world scanned scenes. Each scanned scene contains 2 to 5 identical CAD instances, enabling comprehensive evaluation of point cloud registration methods. The dataset is split into three subsets: 1523 scenes for training, 326 scenes for validation, and an additional 326 scenes for testing. The core task of this dataset is multi-point cloud instance registration.

搜集汇总

背景与挑战

背景概述

Scan2CAD是一个用于3D扫描与CAD模型对齐的数据集，包含97607个关键点对应关系和14225个对象对齐，覆盖1506个扫描场景。数据集支持深度学习研究，特别是3D扫描和CAD模型的对齐任务。