PointAtrousGraph: Deep Hierarchical Encoder-Decoder with Point Atrous Convolution for Unorganized 3D Points

Pan, Liang; Chew, Chee-Meng; Lee, Gim Hee

doi:10.48550/arxiv.1907.09798

“…In this section, we summarize recent research advances with respect to the following three aspects: features by two operations: local pooling [22]- [24] and flexible convolution [25]- [28]. Self-attention often uses linear layers, such as fully-connected (FC) layers and shared multilayer perceptron (shared MLP) layers, which are appropriate for point clouds.…”

Section: Related Workmentioning

confidence: 99%

“…As shown in Fig. 5, RENet follows a hierarchical encoder-decoder architecture by using Edgepreserved Pooling (EP) and Edge-preserved Unpooling (EU) modules [24]. We use an Edge-aware Feature Expansion (EFE) module [5] to expand point features, which generates high-resolution complete point clouds with predicted fine local details.…”

Section: Relational Enhancementmentioning

confidence: 99%

Variational Relational Point Completion Network for Robust 3D Classification

Peng

¹

,

Chen

²

,

Cai

³

et al. 2023

IEEE Trans. Pattern Anal. Mach. Intell.

7

0

View full text Add to dashboard Cite

Real-scanned point clouds are often incomplete due to viewpoint, occlusion, and noise, which hampers 3D geometric modeling and perception. Existing point cloud completion methods tend to generate global shape skeletons and hence lack fine local details. Furthermore, they mostly learn a deterministic partial-to-complete mapping, but overlook structural relations in man-made objects. To tackle these challenges, this paper proposes a variational framework, Variational Relational point Completion network (VRCNet) with two appealing properties: 1) Probabilistic Modeling. In particular, we propose a dual-path architecture to enable principled probabilistic modeling across partial and complete clouds. One path consumes complete point clouds for reconstruction by learning a point VAE. The other path generates complete shapes for partial point clouds, whose embedded distribution is guided by distribution obtained from the reconstruction path during training. 2) Relational Enhancement. Specifically, we carefully design point self-attention kernel and point selective kernel module to exploit relational point features, which refines local shape details conditioned on the coarse completion. In addition, we contribute multi-view partial point cloud datasets (MVP and MVP-40 dataset) containing over 200,000 high-quality scans, which render partial 3D shapes from 26 uniformly distributed camera poses for each 3D CAD model. Extensive experiments demonstrate that VRCNet outperforms state-of-the-art methods on all standard point cloud completion benchmarks. Notably, VRCNet shows great generalizability and robustness on real-world point cloud scans. Moreover, we can achieve robust 3D classification for partial point clouds with the help of VRCNet, which can highly increase classification accuracy. Our project is available at https://paul007pl.github.io/projects/VRCNet.

show abstract

“…PointNet [10] PointNet++ [60] Anand et al [56] Wolf et al [57] Weinmann et al [59] RF_MSSF[58] MS+CU [61] PointSIFT [62] SO-Net [64] RandLA-Net [52] Engelmann et al [63] PAT [66] LSANet [67] PyramNet [68] PointCNN [69] Tangent Convolution [75] DAR-Net [76] A-CNN [70] KPConv [71] ShellNet [77] DPC [72] PointAtrousGraph [74] PointAtrousNet [73] G+RCU [61] RSNet [78] SPLATNet [79] LatticeNet [81] Lawin et al [83] Boulch et al [84] SqueezeSeg [11] SqueezeSegV2[] Gao et al [86] RangeNet++ [87] DeepTemporalSeg [88] LiSeg [89] PointSeg [90] RIU-Net [91] Huang et al [92] SEGCloud [93] FCPN [94] 3DCNN-DQN-RNN[100] SSCN [95] Zhang et al [96] ScanComplete [97] VV-NET [98] VolMap [99] SPG [101] GACNet [102] Jiang et al…”

Section: D Semantic Segmentationmentioning

confidence: 99%

Are We Hungry for 3D LiDAR Data for Semantic Segmentation? A Survey and Experimental Study

Gao¹,

Pan²,

Li³

et al. 2020

Preprint

View full text Add to dashboard Cite

3D LiDAR semantic segmentation is a pivotal task that is widely involved in many applications, such as autonomous driving and robotics. Studies of 3D LiDAR semantic segmentation have recently achieved considerable development, especially in terms of deep learning strategies. However, these studies usually rely heavily on considerable fine annotated data, while point-wise 3D LiDAR datasets are extremely insufficient and expensive to label. The performance limitation caused by the lack of training data is called the data hungry effect. This survey aims to explore whether and how we are hungry for 3D LiDAR data for semantic segmentation. Thus, we first provide an organized review of existing 3D datasets and 3D semantic segmentation methods. Then, we provide an in-depth analysis of three representative datasets and several experiments to evaluate the data hungry effects in different aspects. Efforts to solve data hungry problems are summarized for both 3D LiDAR-focused methods and general-purpose methods. Finally, insightful topics are discussed for future research on data hungry problems and open questions.

show abstract

Variational Relational Point Completion Network

Pan

¹

,

Chen

²

,

Cai³

et al. 2021

2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

View full text Add to dashboard Cite

Real-scanned point clouds are often incomplete due to viewpoint, occlusion, and noise, which hampers 3D geometric modeling and perception. Existing point cloud completion methods tend to generate global shape skeletons and hence lack fine local details. Furthermore, they mostly learn a deterministic partial-to-complete mapping, but overlook structural relations in man-made objects. To tackle these challenges, this paper proposes a variational framework, Variational Relational point Completion network (VRCNet) with two appealing properties: 1) Probabilistic Modeling. In particular, we propose a dual-path architecture to enable principled probabilistic modeling across partial and complete clouds. One path consumes complete point clouds for reconstruction by learning a point VAE. The other path generates complete shapes for partial point clouds, whose embedded distribution is guided by distribution obtained from the reconstruction path during training. 2) Relational Enhancement. Specifically, we carefully design point self-attention kernel and point selective kernel module to exploit relational point features, which refines local shape details conditioned on the coarse completion. In addition, we contribute multi-view partial point cloud datasets (MVP and MVP-40 dataset) containing over 200,000 high-quality scans, which render partial 3D shapes from 26 uniformly distributed camera poses for each 3D CAD model. Extensive experiments demonstrate that VRCNet outperforms state-of-the-art methods on all standard point cloud completion benchmarks. Notably, VRCNet shows great generalizability and robustness on real-world point cloud scans. Moreover, we can achieve robust 3D classification for partial point clouds with the help of VRCNet, which can highly increase classification accuracy. Our project is available at https://paul007pl.github.io/projects/VRCNet.

show abstract

PointAtrousGraph: Deep Hierarchical Encoder-Decoder with Point Atrous Convolution for Unorganized 3D Points

Cited by 4 publications

References 70 publications

Variational Relational Point Completion Network for Robust 3D Classification

Variational Relational Point Completion Network for Robust 3D Classification

Are We Hungry for 3D LiDAR Data for Semantic Segmentation? A Survey and Experimental Study

Variational Relational Point Completion Network

Contact Info

Product

Resources

About