3D Model Classification Based on Contrastive Learning

At present, 3D model classification has been a research hotspot. Massive 3D models not only have diversity in each class, but also have similarities between classes, which seriously affect the classification accuracy of 3D models. We propose a 3D model classification method based on contrastive lear...

Full description

Saved in:
Bibliographic Details
Main Authors: QU Zhongshui, LIU Shan, GAO Yuan, DING Bo
Format: Article
Language:zho
Published: Harbin University of Science and Technology Publications 2025-04-01
Series:Journal of Harbin University of Science and Technology
Subjects:
Online Access:https://hlgxb.hrbust.edu.cn/#/digest?ArticleID=2411
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849706684360949760
author QU Zhongshui
LIU Shan
GAO Yuan
DING Bo
author_facet QU Zhongshui
LIU Shan
GAO Yuan
DING Bo
author_sort QU Zhongshui
collection DOAJ
description At present, 3D model classification has been a research hotspot. Massive 3D models not only have diversity in each class, but also have similarities between classes, which seriously affect the classification accuracy of 3D models. We propose a 3D model classification method based on contrastive learning. In this method, the training is divided into a sample discrimination stage and a classification stage. In the stage of sample discrimination, 3D models of the same category are mutually positive samples, and 3D models of other categories are mutually negative samples. The contrastive loss is used to constrain the sample features, and the positive and negative samples are mapped to the single-center unit hypersphere in the same space to obtain a good semantic representation space of 3D model classification. In addition, in order to capture the correlation between the views and the key areas in the views, a multi head self-attention module and spatial attention module are introduced in the paper. Moreover, the channel attention is added in the multi-head self-attention module to obtain the channel dimension information. In the classification stage, the network model is transferred to the classification task by fine-tuning the network parameters to complete the 3D model classification. The experimental results show that the classification accuracy of the 3D model respectively reaches 99. 4% and 97. 5% on the ModelNet10 and ModelNet40 datasets.
format Article
id doaj-art-a9981bde39ea4fedb759f0751bf88de7
institution DOAJ
issn 1007-2683
language zho
publishDate 2025-04-01
publisher Harbin University of Science and Technology Publications
record_format Article
series Journal of Harbin University of Science and Technology
spelling doaj-art-a9981bde39ea4fedb759f0751bf88de72025-08-20T03:16:07ZzhoHarbin University of Science and Technology PublicationsJournal of Harbin University of Science and Technology1007-26832025-04-013002324110.15938/j.jhust.2025.02.0043D Model Classification Based on Contrastive LearningQU Zhongshui0LIU Shan1GAO Yuan2DING Bo3School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080,ChinaSchool of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080,ChinaSchool of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080,ChinaSchool of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080,ChinaAt present, 3D model classification has been a research hotspot. Massive 3D models not only have diversity in each class, but also have similarities between classes, which seriously affect the classification accuracy of 3D models. We propose a 3D model classification method based on contrastive learning. In this method, the training is divided into a sample discrimination stage and a classification stage. In the stage of sample discrimination, 3D models of the same category are mutually positive samples, and 3D models of other categories are mutually negative samples. The contrastive loss is used to constrain the sample features, and the positive and negative samples are mapped to the single-center unit hypersphere in the same space to obtain a good semantic representation space of 3D model classification. In addition, in order to capture the correlation between the views and the key areas in the views, a multi head self-attention module and spatial attention module are introduced in the paper. Moreover, the channel attention is added in the multi-head self-attention module to obtain the channel dimension information. In the classification stage, the network model is transferred to the classification task by fine-tuning the network parameters to complete the 3D model classification. The experimental results show that the classification accuracy of the 3D model respectively reaches 99. 4% and 97. 5% on the ModelNet10 and ModelNet40 datasets.https://hlgxb.hrbust.edu.cn/#/digest?ArticleID=24113d model classificationcontrastive learningconvolution neural networkattention mechanismtransfer learning
spellingShingle QU Zhongshui
LIU Shan
GAO Yuan
DING Bo
3D Model Classification Based on Contrastive Learning
Journal of Harbin University of Science and Technology
3d model classification
contrastive learning
convolution neural network
attention mechanism
transfer learning
title 3D Model Classification Based on Contrastive Learning
title_full 3D Model Classification Based on Contrastive Learning
title_fullStr 3D Model Classification Based on Contrastive Learning
title_full_unstemmed 3D Model Classification Based on Contrastive Learning
title_short 3D Model Classification Based on Contrastive Learning
title_sort 3d model classification based on contrastive learning
topic 3d model classification
contrastive learning
convolution neural network
attention mechanism
transfer learning
url https://hlgxb.hrbust.edu.cn/#/digest?ArticleID=2411
work_keys_str_mv AT quzhongshui 3dmodelclassificationbasedoncontrastivelearning
AT liushan 3dmodelclassificationbasedoncontrastivelearning
AT gaoyuan 3dmodelclassificationbasedoncontrastivelearning
AT dingbo 3dmodelclassificationbasedoncontrastivelearning