Convolution in the Cloud: Learning Deformable Kernels in 3D Graph Convolution Networks for Point Cloud Analysis

Lin, Zhihao; Huang, Shengyu; Wang, Yu-Chiang Frank

doi:10.1109/cvpr42600.2020.00187

Cited by 207 publications

(102 citation statements)

References 23 publications

Supporting

Mentioning

102

Contrasting

Order By: Relevance

“…PointAtrousGraph [27] enlarges the receptive fields of filters while densely learning multiscale point features. 3DGCN [22] learns deformable 3D kernel by learning directional information within local receptive fields to guarantee scale invariance. KCNet [37] mines local points with similar geometric structures to capture neighborhood information.…”

Section: Point-based Methodsmentioning

confidence: 99%

MeshNet++: A Network with a Face

Singh

Sheshappanavar

Kambhamettu

2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

a) Decimated Mesh (b) MeshNet Point Coverage (c) MeshNet++ Point CoverageFigure 1: Robustness to decimation by MeshNet++; (a) Mesh Decimated to 1024 faces; (b) Points captured by MeshNet while generating structural and spatial features; (c) Points captured by MeshNet++ while generating only structural features. It is evident that MeshNet++ captures the original shape of a mesh and is robust to the shortcomings of mesh decimation.

show abstract

Section: Point-based Methodsmentioning

confidence: 99%

MeshNet++: A Network with a Face

Singh

Sheshappanavar

Kambhamettu

2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

show abstract

“…However, an increase in the number of SM layer does not benefit the network. 1.48M 91.9 3DmFV [42] 45.77M 91.6 KPConv [19] 14.3M 92.9 DGCNN [17] 1.81M 92.9 PointCNN [37] 0.6M 92.2 3D-GCN [43] 0.89M 92.1 Ours 1.7M 93.2…”

Section: Methodsmentioning

confidence: 99%

Semantic Feature Mining for 3D Object Classification and Segmentation

Zhao

Premebida

et al. 2021

2021 IEEE International Conference on Robotics and Automation (ICRA)

View full text Add to dashboard Cite

Deep learning on 3D point clouds has drawn much attention, due to its large variety of applications in intelligent perception for automated and robotic systems. Unlike structured 2D images, it is challenging to extract features and implement convolutional networks over these unordered points. Although a number of previous works achieved high accuracies for point cloud recognition, they tend to process local point information in such a way that semantic information is not fully encoded. In this paper, we propose a deep neural network for 3D point cloud processing that utilizes effective feature aggregation methods emphasizing both generalizability and relevance. In particular, our method uses fixed-radius grouping for pooling layers and spherical kernel convolution for semantics mining. To address the issue of gradient degradation and memory consumption of a deep network, a parallel feature feed-forward mechanism and bottleneck layers are implemented to reduce the number of parameters. Experiments show that our algorithm achieves state-of-the-art results and competitive accuracy in both classification and part segmentation while maintaining an efficient architecture.

show abstract

“…Although there are many network architectures that directly process point cloud [26,27,45], most of the architectures calculate on point coordinates, which means their networks are sensitive to point clouds shift and size variation [18]. This decreases the pose estimation accuracy.…”

Section: Shape-based Networkmentioning

confidence: 99%

“…Therefore, the performance of these methods is not guaranteed with limited training examples. To overcome this issue, we propose a 3D graph convolution (3DGC) autoencoder [18] to effectively learn the category-level pose feature via observed points reconstruction of different objects instead of uniform shape mapping. We further propose an online box-cage based 3D data augmentation mechanism to reduce the dependencies of labeled data.…”

Section: Introductionmentioning

confidence: 99%

FS-Net: Fast Shape-based Network for Category-Level 6D Object Pose Estimation with Decoupled Rotation Mechanism

Chen

Jia

Chang

et al. 2021

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

128

View full text Add to dashboard Cite

In this paper, we focus on category-level 6D pose and size estimation from a monocular RGB-D image. Previous methods suffer from inefficient category-level pose feature extraction, which leads to low accuracy and inference speed. To tackle this problem, we propose a fast shapebased network (FS-Net) with efficient category-level feature extraction for 6D pose estimation. First, we design an orientation aware autoencoder with 3D graph convolution for latent feature extraction. Thanks to the shift and scaleinvariance properties of 3D graph convolution, the learned latent feature is insensitive to point shift and object size. Then, to efficiently decode category-level rotation information from the latent feature, we propose a novel decoupled rotation mechanism that employs two decoders to complementarily access the rotation information. For translation and size, we estimate them by two residuals: the difference between the mean of object points and ground truth translation, and the difference between the mean size of the category and ground truth size, respectively. Finally, to increase the generalization ability of the FS-Net, we propose an online box-cage based 3D deformation mechanism to augment the training data. Extensive experiments on two benchmark datasets show that the proposed method achieves state-ofthe-art performance in both category-and instance-level 6D object pose estimation. Especially in category-level pose estimation, without extra synthetic data, our method outperforms existing methods by 6.3% on the NOCS-REAL dataset 1 .

show abstract

Convolution in the Cloud: Learning Deformable Kernels in 3D Graph Convolution Networks for Point Cloud Analysis

Cited by 207 publications

References 23 publications

MeshNet++: A Network with a Face

MeshNet++: A Network with a Face

Semantic Feature Mining for 3D Object Classification and Segmentation

FS-Net: Fast Shape-based Network for Category-Level 6D Object Pose Estimation with Decoupled Rotation Mechanism

Contact Info

Product

Resources

About