CADNET

我们介绍了 CADNET 数据集，它是 43 个类别的 3,317 个 3D 工程模型的注释集合。由于大型注释数据集的可用性以及 GPU 形式的足够计算能力，最近提出了许多基于深度学习的对象分类解决方案，特别是在图像和图形模型领域。然而，针对 CAD 模型的功能分类任务提出的解决方案很少。因此，对于这项研究，CAD 模型已从工程形状基准 (ESB)、国家设计存储库 (NDR) 中收集，并使用建模软件创建的新模型进行扩充，以形成数据集 - “CADNET”。We introduce the CADNET dataset, which is a collection of annotated 3D engineering models totaling 3,317 instances across 43 categories. Given the availability of large annotated datasets and sufficient computational power in the form of GPUs, numerous deep learning-based object classification solutions have been proposed recently, particularly in the fields of image and graphical models. However, very few solutions have been put forward for the functional classification task of CAD models. Therefore, for this study, CAD models were collected from the Engineering Shape Benchmark (ESB) and the National Design Repository (NDR), and augmented with new models created using modeling software to form the "CADNET" dataset.

搜集汇总

背景与挑战

背景概述

CADNET是一个专注于CAD模型功能分类的3D工程模型数据集，包含43个类别的3,317个模型，数据来源于ESB、NDR和新创建模型。该数据集由印度理工学院马德拉斯分校于2021年发布，旨在支持基于深度学习的CAD分类研究，填补了该领域标注数据不足的空白。