AttPNet: Attention-Based Deep Neural Network for 3D Point Set Analysis
KAUST DepartmentComputational Bioscience Research Center (CBRC)
Computer Science Program
Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Structural and Functional Bioinformatics Group
Permanent link to this recordhttp://hdl.handle.net/10754/665326
MetadataShow full item record
AbstractPoint set is a major type of 3D structure representation format characterized by its data availability and compactness. Most former deep learning-based point set models pay equal attention to different point set regions and channels, thus having limited ability in focusing on small regions and specific channels that are important for characterizing the object of interest. In this paper, we introduce a novel model named Attention-based Point Network (AttPNet). It uses attention mechanism for both global feature masking and channel weighting to focus on characteristic regions and channels. There are two branches in our model. The first branch calculates an attention mask for every point. The second branch uses convolution layers to abstract global features from point sets, where channel attention block is adapted to focus on important channels. Evaluations on the ModelNet40 benchmark dataset show that our model outperforms the existing best model in classification tasks by 0.7% without voting. In addition, experiments on augmented data demonstrate that our model is robust to rotational perturbations and missing points. We also design a Electron Cryo-Tomography (ECT) point cloud dataset and further demonstrate our model’s ability in dealing with fine-grained structures on the ECT dataset.
CitationYang, Y., Ma, Y., Zhang, J., Gao, X., & Xu, M. (2020). AttPNet: Attention-Based Deep Neural Network for 3D Point Set Analysis. Sensors, 20(19), 5455. doi:10.3390/s20195455
SponsorsThis work was supported in part by U.S. National Institutes of Health (NIH) grant P41GM103712, R01GM134020, and K01MH123896, U.S. National Science Foundation (NSF) grant DBI-1949629 and IIS-2007595. Mark Foundation for Cancer Research grant 19-044-ASP, King Abdullah University of Science and Technology (KAUST) Office of Sponsored Research (OSR) under Award No. URF/1/2602-01 and URF/1/3007-01.
Except where otherwise noted, this item's license is described as This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
- A Novel Point Cloud Encoding Method Based on Local Information for 3D Classification and Segmentation.
- Authors: Song Y, Gao L, Li X, Shen W
- Issue date: 2020 Apr 28
- Fine-Grained 3D Shape Classification With Hierarchical Part-View Attention.
- Authors: Liu X, Han Z, Liu YS, Zwicker M
- Issue date: 2021
- A Deep Learning Method for 3D Object Classification and Retrieval Using the Global Point Signature Plus and Deep Wide Residual Network.
- Authors: Hoang L, Lee SH, Kwon KR
- Issue date: 2021 Apr 9
- Generating Point Cloud from Measurements and Shapes Based on Convolutional Neural Network: An Application for Building 3D Human Model.
- Authors: Nguyen MT, Dang TV, Tran Thi MK, The Bao P
- Issue date: 2019
- TIF-Reg: Point Cloud Registration with Transform-Invariant Features in SE(3).
- Authors: Chen B, Chen H, Song B, Gong G
- Issue date: 2021 Aug 27