Mesh Convolutional Restricted Boltzmann Machines for Unsupervised Learning of Features with Structure Preservation on 3D Meshes

UM > Faculty of Science and Technology > DEPARTMENT OF COMPUTER AND INFORMATION SCIENCE

Status	已發表Published
	Mesh Convolutional Restricted Boltzmann Machines for Unsupervised Learning of Features with Structure Preservation on 3D Meshes
	Han, Z.; Liu, Z.; Han, J.; Vong, C. M.; Bu, S.; Chen, C. L.
	2017-10-01
Source Publication	IEEE Transactions on Neural Networks and Learning Systems (SCI-E)
ISSN	2162-237X
Pages	2268-2281
Abstract	Discriminative features of 3D meshes are significant to many 3D shape analysis tasks. However, handcrafted descriptors and traditional unsupervised 3D feature learning methods suffer from several significant weaknesses: i) the extensive human intervention is involved; ii) the local and global structure information of 3D meshes cannot be preserved, which is in fact an important source of discriminability; iii) the irregular vertex topology and arbitrary resolution of 3D meshes do not allow the direct application of the popular deep learning models; iv) the orientation is ambiguous on the mesh surface; v) the effect of rigid and non-rigid transformations on 3D meshes cannot be eliminated. As a remedy, we propose a deep learning model with a novel irregular model structure, called Mesh Convolutional Restricted Boltzmann Machines (MCRBM). MCRBM aims to simultaneously learn structure-preserving local and global features from a novel raw representation, Local Function Energy Distribution (LFED). In addition, multiple MCRBMs can be stacked into a deeper model, called Mesh Convolutional Deep Belief Networks (MCDBN). MCDBN employs a novel Local Structure Preserving Convolution (LSPC) strategy to convolve the geometry and the local structure learned by the lower MCRBM to the upper MCRBM.
Keyword	3D mesh Laplace-Beltrami operator Mesh Convolutional Restricted Boltzman Machines Mesh Convolutional Deep Belief Networks
Language	英語English
The Source to Article	PB_Publication
PUB ID	20348
Document Type	Journal article
Collection	DEPARTMENT OF COMPUTER AND INFORMATION SCIENCE
Corresponding Author	Liu, Z.
Recommended Citation GB/T 7714	Han, Z.,Liu, Z.,Han, J.,et al. Mesh Convolutional Restricted Boltzmann Machines for Unsupervised Learning of Features with Structure Preservation on 3D Meshes[J]. IEEE Transactions on Neural Networks and Learning Systems (SCI-E), 2017, 2268-2281.
APA	Han, Z.., Liu, Z.., Han, J.., Vong, C. M.., Bu, S.., & Chen, C. L. (2017). Mesh Convolutional Restricted Boltzmann Machines for Unsupervised Learning of Features with Structure Preservation on 3D Meshes. IEEE Transactions on Neural Networks and Learning Systems (SCI-E), 2268-2281.
MLA	Han, Z.,et al."Mesh Convolutional Restricted Boltzmann Machines for Unsupervised Learning of Features with Structure Preservation on 3D Meshes".IEEE Transactions on Neural Networks and Learning Systems (SCI-E) (2017):2268-2281.

Files in This Item:
There are no files associated with this item.

If you have any objections to this item, please fill out the form below and the administrator will contact you as soon as possible.
Content:
Email：	*
Affiliation No.
Verification Code:	Refresh

Any comments and suggestions are welcomed.
Title:	*
Content:
Email：	*
Verification Code:	Refresh