|
Journal of Zhejiang University SCIENCE C
ISSN 1869-1951(Print), 1869-196x(Online), Monthly
2014 Vol.15 No.2 P.91-106
A deep learning approach to the classification of 3D CAD models
Abstract: Model classification is essential to the management and reuse of 3D CAD models. Manual model classification is laborious and error prone. At the same time, the automatic classification methods are scarce due to the intrinsic complexity of 3D CAD models. In this paper, we propose an automatic 3D CAD model classification approach based on deep neural networks. According to prior knowledge of the CAD domain, features are selected and extracted from 3D CAD models first, and then preprocessed as high dimensional input vectors for category recognition. By analogy with the thinking process of engineers, a deep neural network classifier for 3D CAD models is constructed with the aid of deep learning techniques. To obtain an optimal solution, multiple strategies are appropriately chosen and applied in the training phase, which makes our classifier achieve better performance. We demonstrate the efficiency and effectiveness of our approach through experiments on 3D CAD model datasets.
Key words: CAD model classification, Design reuse, Machine learning, Neural network
创新要点:利用深度学习技术,模拟工程师人工识别三维零件的三个主要环节,构建了CAD模型的自动分类器;采用多种行之有效的训练策略,得到具有较好泛化性能的分类器。
研究方法:基于深度学习的三维CAD模型分类方法的主要步骤为:从制造企业获取足够的样本数据,构建三维CAD模型数据集;分析CAD领域的特点,选取合适的三维模型描述符进行特征提取,并处理为输入向量;利用深度学习技术模拟工程师人工识别工程零件的过程,设计分类器的拓扑结构并选择相应的组成元素,构建深层神经网络分类器;采用多种实用的训练技巧,使训练得到的分类器的自由参数取值尽可能逼近全局最优解,避免过拟合。
重要结论:实验数据表明,将深度学习技术与CAD领域知识结合,能够取得满意效果。深层神经网络分类器在测试集上的分类误差率低,足以满足工业生产的需求。
关键词组:
References:
Open peer comments: Debate/Discuss/Question/Opinion
<1>
editor@No address<No mail>
2014-01-29 19:26:10
Reviewer: This paper proposed a method for automatically classifying CAD models ba
This work is the first one to my knowledge to apply the technique of deep learning in the classification of 3D models. The algorithm is technically sound and the results are good.
Deep learning has been widely studied and used in image community. I am curious why this technique is used in 3D content until now. And it is nice to see that this technique does work well for 3D shape classification.
DOI:
10.1631/jzus.C1300185
CLC number:
TP391.72
Download Full Text:
Downloaded:
16062
Download summary:
<Click Here>Downloaded:
2979Clicked:
12075
Cited:
6
On-line Access:
2024-08-27
Received:
2023-10-17
Revision Accepted:
2024-05-08
Crosschecked:
2014-01-15