Filters








157 Hits in 3.3 sec

Large-Scale 3D Shape Retrieval from ShapeNet Core55 [article]

Manolis Savva, Fisher Yu, Hao Su, Masaki Aono, Baoquan Chen, Daniel Cohen-Or, Weihong Deng, Hang Su, Song Bai, Xiang Bai, Noa Fish, Jiajie Han (+9 others)
2016 Eurographics Workshop on 3D Object Retrieval, EG 3DOR  
This track aims to provide a benchmark to evaluate large-scale shape retrieval based on the ShapeNet dataset.  ...  We use ShapeNet Core55, which provides more than 50 thousands models over 55 common categories in total for training and evaluating several algorithms.  ...  The covariance matrix lies not on the Euclidean space, but on the Riemannian manifold of symmetric positive semi-define matrices.  ... 
doi:10.2312/3dor.20161092 fatcat:6yoekzs7zfchnkztrmnicxy3cq

NPTC-net: Narrow-Band Parallel Transport Convolutional Neural Network on Point Clouds [article]

Pengfei Jin, Tianhao Lai, Rongjie Lai, Bin Dong
2021 arXiv   pre-print
Designing appropriate convolution neural networks on manifold-structured point clouds can inherit and empower recent advances of CNNs to analyzing and processing point cloud data.  ...  With that, we further propose a deep convolutional neural network based on NPTC (called NPTC-net) for point cloud classification and segmentation.  ...  One of the main challenges of proposing geometric meaningful convolution on manifolds and point clouds (a discrete form of manifolds) is to define an analogy of the Euclidean translation x − y on the non-Euclidean  ... 
arXiv:1905.12218v3 fatcat:ogcgub3ebndgbmlpwab6sd4tkq

SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters [chapter]

Yifan Xu, Tianqi Fan, Mingye Xu, Long Zeng, Yu Qiao
2018 Lecture Notes in Computer Science  
Deep neural networks have enjoyed remarkable success for various vision tasks, however it remains challenging to apply CNNs to domains lacking a regular underlying structures such as 3D point clouds.  ...  Experiments on ModelNet40 demonstrate that SpiderCNN achieves stateof-the-art accuracy 92.4% on standard benchmarks, and shows competitive performance on segmentation task.  ...  Various studies are devoted to making convolution neural networks applicable for learning on non-Euclidean domains such as graphs or manifolds by trying to generalize the definition of convolution to functions  ... 
doi:10.1007/978-3-030-01237-3_6 fatcat:ewjesxzc5zcnrmdg42b7lhxjoa

SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters [article]

Yifan Xu, Tianqi Fan, Mingye Xu, Long Zeng, Yu Qiao
2018 arXiv   pre-print
Deep neural networks have enjoyed remarkable success for various vision tasks, however it remains challenging to apply CNNs to domains lacking a regular underlying structures such as 3D point clouds.  ...  Experiments on ModelNet40 demonstrate that SpiderCNN achieves state-of-the-art accuracy 92.4% on standard benchmarks, and shows competitive performance on segmentation task.  ...  Various studies are devoted to making convolution neural networks applicable for learning on non-Euclidean domains such as graphs or manifolds by trying to generalize the definition of convolution to functions  ... 
arXiv:1803.11527v3 fatcat:tsb3m6iamrconbyn3yfhs6l4rq

LaplacianNet: Learning on 3D Meshes with Laplacian Encoding and Pooling [article]

Yi-Ling Qiao, Lin Gao, Jie Yang, Paul L. Rosin, Yu-Kun Lai, Xilin Chen
2019 arXiv   pre-print
Our network architecture is flexible enough to be used on meshes with different numbers of vertices.  ...  We conduct several experiments including shape segmentation and classification, and our LaplacianNet outperforms state-of-the-art algorithms for these tasks on ShapeNet and COSEG datasets.  ...  We therefore extend this ideology into non-Euclidean space of 2-manifolds. In our design as in Figure 1 , the basic network structure involves consecutive Mesh Pooling Blocks (MPBs).  ... 
arXiv:1910.14063v1 fatcat:sqzdkvvxrnerhaxzn3z2kpgkum

Geodesic-HOF: 3D Reconstruction Without Cutting Corners [article]

Ziyun Wang, Eric A. Mitchell, Volkan Isler, Daniel D. Lee
2020 arXiv   pre-print
on the object.  ...  To address this issue, we propose learning an image-conditioned mapping function from a canonical sampling domain to a high dimensional space where the Euclidean distance is equal to the geodesic distance  ...  With the success of Convolutional Neural Network (CNN) in processing image data, it is natural to extend it to 3D.  ... 
arXiv:2006.07981v1 fatcat:msyxgquzkjbx7fv77vawjkvaey

IntrA: 3D Intracranial Aneurysm Dataset for Deep Learning [article]

Xi Yang, Ding Xia, Taichi Kin, Takeo Igarashi
2020 arXiv   pre-print
We provide a large-scale benchmark of classification and part segmentation by testing state-of-the-art networks.  ...  Objects with arbitrary shapes are ubiquitous, and a non-Euclidean manifold reveals more critical information than using Euclidean geometry, like complex typologies of brain tissues in neuroscience [7]  ...  The methods based on projected view or voxel are implemented conveniently using similar structures with 2D convolutional neural networks (CNNs).  ... 
arXiv:2003.02920v2 fatcat:ecsico6livgwpbmlqgjyuijvwe

3D Pick & Mix: Object Part Blending in Joint Shape and Image Manifolds [article]

Adrian Penate-Sanchez, Lourdes Agapito
2018 arXiv   pre-print
Many applications could benefit from such rich queries, users could browse through catalogues of furniture and pick and mix parts, combining for example the legs of a chair from one shop and the armrests  ...  The next step is to train a deep neural network that can map RGB images onto each manifold by regressing the coordinates on each part manifold directly from RGB inputs.  ...  The Euclidean loss is chosen since the part shape manifolds are themselves Euclidean spaces.  ... 
arXiv:1811.01068v1 fatcat:ylj4bgefajfcxpluasszn7mgte

A Survey on Deep Geometry Learning: From a Representation Perspective [article]

Yun-Peng Xiao, Yu-Kun Lai, Fang-Lue Zhang, Chunpeng Li, Lin Gao
2020 arXiv   pre-print
Researchers have now achieved great success on dealing with 2D images using deep learning. In recent years, 3D computer vision and Geometry Deep Learning gain more and more attention.  ...  However, the performance for different applications largely depends on the representation used, and there is no unique representation that works well for all applications.  ...  Several works are designed in 2-manifolds with a series of refined CNN operators to adapt to this non-Euclidean space.  ... 
arXiv:2002.07995v2 fatcat:pustwlu5freypnccfrculkqvei

A-CNN: Annularly Convolutional Neural Networks on Point Clouds [article]

Artem Komarichev, Zichun Zhong, Jing Hua
2019 arXiv   pre-print
This paper presents a new method to define and compute convolution directly on 3D point clouds by the proposed annular convolution.  ...  We apply it to the developed hierarchical neural networks for object classification, part segmentation, and semantic segmentation in large-scale scenes.  ...  Meanwhile, several approaches have been proposed to develop convolutional networks on 2D manifolds [22, 4, 24, 41] .  ... 
arXiv:1904.08017v1 fatcat:itqbfcvjyzgtplapzpnr5m3we4

A survey on deep geometry learning: From a representation perspective

Yun-Peng Xiao, Yu-Kun Lai, Fang-Lue Zhang, Chunpeng Li, Lin Gao
2020 Computational Visual Media  
The performance achieved in different applications largely depends on the representation used, and there is no unique representation that works well for all applications.  ...  Several works have been designed using 2-manifolds with a series of refined CNN operators adapted to such non-Euclidean spaces.  ...  One of the earliest methods to apply deep neural networks to volumetric representations, 3D ShapeNets, was proposed by Wu et al. [13] in 2015.  ... 
doi:10.1007/s41095-020-0174-8 fatcat:kpoynaixq5esbek63bovybisfa

Meshing Point Clouds with Predicted Intrinsic-Extrinsic Ratio Guidance [article]

Minghua Liu, Xiaoshuai Zhang, Hao Su
2020 arXiv   pre-print
We learn to predict this surrogate using a deep point cloud network and then feed it to an efficient post-processing module for high-quality mesh generation.  ...  We demonstrate that our method can not only preserve details, handle ambiguous structures, but also possess strong generalizability to unseen categories by experiments on synthetic and real data.  ...  Since meshes in ShapeNet are not manifolds, it's not trivial to calculate point normals with consistent directions, but many algorithms rely on correct normals.  ... 
arXiv:2007.09267v2 fatcat:onxslft3y5cojmz6mm57oo7xqi

Orientation-Encoding CNN for Point Cloud Classification and Segmentation

Hongbin Lin, Wu Zheng, Xiuping Peng
2021 Machine Learning and Knowledge Extraction  
However, there are still tough challenges in applying them to convolutional neural networks due to the lack of a potential rule structure of point clouds.  ...  Therefore, by taking the original point clouds as the input data, this paper proposes an orientation-encoding (OE) convolutional module and designs a convolutional neural network for effectively extracting  ...  Spatial Domain The GeodesicCNN [17] is a generalization of the convolution network paradigm to non-Euclidean manifolds.  ... 
doi:10.3390/make3030031 fatcat:jj2jmeu42fhavgh5ebj5y2xyxq

Learning to Reconstruct Symmetric Shapes using Planar Parameterization of 3D Surface

Hardik Jain, Manuel Wollhaf, Olaf Hellwich
2019 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW)  
This representation is encoded with surface information to generate 2D geometry images, which can be conveniently learned using traditional deep neural networks without additional overhead.  ...  Prior knowledge encoded in trained networks has proven to be effective in generating images. Based on a similar paradigm, various methods were proposed to generate 3D shape from images.  ...  Meshes not following Equation 1 are often referred to as non-manifold. These meshes have non-manifoldness in at least one of the vertices or edges.  ... 
doi:10.1109/iccvw.2019.00508 dblp:conf/iccvw/JainWH19 fatcat:c4t5osl2b5bkxdyzldh3xuhfpy

Faster Dynamic Graph CNN : Faster Deep Learning on 3D Point Cloud Data

Jinseok Hong, Keeyoung Kim, Hongchul Lee
2020 IEEE Access  
A geodesic CNN (GCNN) [58] is a non-Euclidean deep learning algorithm that uses a spectral filter instead of a spatial filter.  ...  , such as graphics and manifolds.  ... 
doi:10.1109/access.2020.3023423 fatcat:g3cvna7zp5eynpcsacbri52lve
« Previous Showing results 1 — 15 out of 157 results