Linked Dynamic Graph CNN: Learning on Point Cloud via Linking Hierarchical Features

Zhang, Kuangen; Hao, Ming; Wang, Jing; de Silva, Clarence W.; Fu, Chenglong

doi:10.48550/arxiv.1904.10014

Cited by 51 publications

(66 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These models are computationally efficient due to their excellent memory locality, but loss of the inevitable information degrades the fine-grained localization accuracy [27,20]. Instead of voxelization, developing a neutral network that consumes directly on point clouds is possible [22,47,10,26,41,31,34,46,43]. Although these point-based models naturally preserve accuracy of point location, they are usually computationally intensive.…”

Section: Point Cloud Learningmentioning

confidence: 99%

“…Input OA Latency OA<92.5 PointNet [22] 16×1024 89.2 13.6ms PointNet++ [24] 16×1024 91.9 35.3ms SO-Net [13] 8×2048 90.9 − PointGrid [11] 16×1021 92.0 − SpiderCNN [39] 8×1024 92.4 82.6ms PointCNN [15] 16×1024 92.2 221.2ms PointWeb [48] 16×1024 92.3 − PVCNN [20] 16×1024 92.4 24.2ms OA>92.5 KPConv [28] 16×6500 92.9 120.5ms DGCNN [34] 16×1024 92.9 85.8ms LDGCNN [46] 16×1024 92. Implementation details.…”

Section: Modelmentioning

confidence: 99%

See 1 more Smart Citation

PatchFormer: An Efficient Point Transformer with Patch Attention

Zhong-yu¹,

Wan²,

Shen³

et al. 2021

Preprint

View full text Add to dashboard Cite

The point cloud learning community is witnesses a modeling shift from CNNs to Transformers, where pure Transformer architectures have achieved top accuracy on the major learning benchmarks. However, existing point Transformers are computationally expensive since they need to generate a large attention map, which has quadratic complexity (both in space and time) with respect to input size. To solve this shortcoming, we introduce Patch ATtention (PAT) to adaptively learn a much smaller set of bases upon which the attention maps are computed. By a weighted summation upon these bases, PAT not only captures the global shape context but also achieves linear complexity to input size. In addition, we propose a lightweight Multi-Scale aTtention (MST) block to build attentions among features of different scales, providing the model with multi-scale features. Equipped with the PAT and MST, we construct our neural architecture called PatchFormer that integrates both modules into a joint framework for point cloud learning. Extensive experiments demonstrate that our network achieves comparable accuracy on general point cloud learning tasks with 9.2× speed-up than previous point Transformers.

show abstract

Section: Point Cloud Learningmentioning

confidence: 99%

Section: Modelmentioning

confidence: 99%

PatchFormer: An Efficient Point Transformer with Patch Attention

Zhong-yu¹,

Wan²,

Shen³

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The symbol "-" indicates that the results are not available from the references. PointNet++ [12] 90.7% ---SRN-PointNet++ [24] 91.5% ---PointASNL [25] 93.2% -95.9% -Convolution-based PointConv [26] 92.5% ---Methods A-CNN [27] 92.6% 90.3% 95.5% 95.3% SFCNN [28] 92.3% ---InterpCNN [29] 93.0% ---ConvPoint [30] 91.8% 88.5% --Graph-based ECC [23] 87.4% 83.2% 90.8% 90.0% Methods DGCNN [22] 92.2% 90.2% --LDGCNN [31] 92.9% 90.3% --Hassani et al [32] 89.1% ---DPAM [33] 91.9% 89.9% 94.6% 94.3% KCNet [34] 91.0% -94.4% -ClusterNet [35] 87.1% ---RGCNN [36] 90.5% 87.3% --LocalSpecGCN [37] 92.1% ---PointGCN [38] 89.5% 86.1% 91.9% 91.6% 3DTI-Net [39] 91.7% ---Grid-GCN [40] 93…”

Section: ) Hierarchical Prediction Architecturementioning

confidence: 99%

Structure-Aware Multi-Hop Graph Convolution for Graph Neural Networks

Li¹,

Tanaka²

2021

Preprint

View full text Add to dashboard Cite

In this paper, we propose a spatial graph convolution (GC) to classify signals on a graph. Existing GC methods are limited to using the structural information in the feature space. Additionally, the single step of GCs only uses features on the one-hop neighboring nodes from the target node. In this paper, we propose two methods to improve the performance of GCs: 1) Utilizing structural information in the feature space, and 2) exploiting the multi-hop information in one GC step. In the first method, we define three structural features in the feature space: feature angle, feature distance, and relational embedding. The second method aggregates the node-wise features of multi-hop neighbors in a GC. Both methods can be simultaneously used. We also propose graph neural networks (GNNs) integrating the proposed GC for classifying nodes in 3D point clouds and citation networks. In experiments, the proposed GNNs exhibited a higher classification accuracy than existing methods.

show abstract

“…Finally, graph-based approaches create a k-nearest neighbor-or radius-graph from the input set and apply graph convolutions [65,73,87,44,64,17,9,68,21,72,88,54]. DGCNN [73] introduces the EdgeConv operator and a dynamic graph component, which reconnects the k-nearest neighbor graph inside the network.…”

Section: Related Workmentioning

confidence: 99%

DeltaConv: Anisotropic Operators for Geometric Deep Learning on Point Clouds

Wiersma,

Nasikun,

Eisemann

et al. 2021

Preprint

View full text Add to dashboard Cite

Learning from 3D point-cloud data has rapidly gained momentum, motivated by the success of deep learning on images and the increased availability of 3D data. In this paper, we aim to construct anisotropic convolutions that work directly on the surface derived from a point cloud. This is challenging because of the lack of a global coordinate system for tangential directions on surfaces. We introduce a new convolution operator called DeltaConv, which combines geometric operators from exterior calculus to enable the construction of anisotropic filters on point clouds. Because these operators are defined on scalar-and vectorfields, we separate the network into a scalar-and a vectorstream, which are connected by the operators. The vector stream enables the network to explicitly represent, evaluate, and process directional information. Our convolutions are robust and simple to implement and show improved accuracy compared to state-of-the-art approaches on several benchmarks, while also speeding up training and inference.

show abstract

Linked Dynamic Graph CNN: Learning on Point Cloud via Linking Hierarchical Features

Cited by 51 publications

References 24 publications

PatchFormer: An Efficient Point Transformer with Patch Attention

PatchFormer: An Efficient Point Transformer with Patch Attention

Structure-Aware Multi-Hop Graph Convolution for Graph Neural Networks

DeltaConv: Anisotropic Operators for Geometric Deep Learning on Point Clouds

Contact Info

Product

Resources

About