3D2SeqViews: Aggregating Sequential Views for 3D Global Feature Learning by CNN with Hierarchical Attention Aggregation
Han,Zhizhong1,6; Lu,Honglei1; Liu,Zhenbao2; Vong,Chi Man3; Liu,Yu Shen1,7; Zwicker,Matthias4; Han,Junwei2; Philip Chen,C. L.5
Source PublicationIEEE Transactions on Image Processing
AbstractLearning 3D global features by aggregating multiple views is important. Pooling is widely used to aggregate views in deep learning models. However, pooling disregards a lot of content information within views and the spatial relationship among the views, which limits the discriminability of learned features. To resolve this issue, 3D to Sequential Views (3D2SeqViews) is proposed to more effectively aggregate the sequential views using convolutional neural networks with a novel hierarchical attention aggregation. Specifically, the content information within each view is first encoded. Then, the encoded view content information and the sequential spatiality among the views are simultaneously aggregated by the hierarchical attention aggregation, where view-level attention and class-level attention are proposed to hierarchically weight sequential views and shape classes. View-level attention is learned to indicate how much attention is paid to each view by each shape class, which subsequently weights sequential views through a novel recursive view integration. Recursive view integration learns the semantic meaning of view sequence, which is robust to the first view position. Furthermore, class-level attention is introduced to describe how much attention is paid to each shape class, which innovatively employs the discriminative ability of the fine-tuned network. 3D2SeqViews learns more discriminative features than the state-of-the-art, which leads to the outperforming results in shape classification and retrieval under three large-scale benchmarks.
Keyword3D global feature learning CNN hierarchical attention aggregation sequential views view aggregation
URLView the original
Scopus ID2-s2.0-85067824808
Fulltext Access
Citation statistics
Cited Times [WOS]:30   [WOS Record]     [Related Records in WOS]
Document TypeJournal article
CollectionUniversity of Macau
Affiliation1.School of Software,Tsinghua University,Beijing,100084,China
2.School of Aeronautics,Northwestern Polytechnical University,Xi'an,710072,China
3.Department of Computer and Information Science,University of Macau,Macau,99999,Macao
4.University of Maryland,College Park,20737,United States
5.Faculty of Science and Technology,University of Macau,Macau,99999,Macao
6.Department of Computer Science,University of Maryland,College Park,20737,United States
7.Beijing National Research Center for Information Science and Technology,China
Recommended Citation
GB/T 7714
Han,Zhizhong,Lu,Honglei,Liu,Zhenbao,et al. 3D2SeqViews: Aggregating Sequential Views for 3D Global Feature Learning by CNN with Hierarchical Attention Aggregation[J]. IEEE Transactions on Image Processing,2019,28(8):3986-3999.
APA Han,Zhizhong,Lu,Honglei,Liu,Zhenbao,Vong,Chi Man,Liu,Yu Shen,Zwicker,Matthias,Han,Junwei,&Philip Chen,C. L..(2019).3D2SeqViews: Aggregating Sequential Views for 3D Global Feature Learning by CNN with Hierarchical Attention Aggregation.IEEE Transactions on Image Processing,28(8),3986-3999.
MLA Han,Zhizhong,et al."3D2SeqViews: Aggregating Sequential Views for 3D Global Feature Learning by CNN with Hierarchical Attention Aggregation".IEEE Transactions on Image Processing 28.8(2019):3986-3999.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Han,Zhizhong]'s Articles
[Lu,Honglei]'s Articles
[Liu,Zhenbao]'s Articles
Baidu academic
Similar articles in Baidu academic
[Han,Zhizhong]'s Articles
[Lu,Honglei]'s Articles
[Liu,Zhenbao]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Han,Zhizhong]'s Articles
[Lu,Honglei]'s Articles
[Liu,Zhenbao]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.