Pointwise Convolutional Neural Networks

Hua, Binh-Son; Tran, Minh-Khoi; Yeung, Sai-Kit

doi:10.1109/cvpr.2018.00109

Cited by 485 publications

(227 citation statements)

References 38 publications

Supporting

Mentioning

196

Contrasting

Unclassified

Order By: Relevance

“…Notice that we replace K-ring neighborhoods with K-nearest neighbors for RoSt and Jets to support point cloud input. Also, 7 machine learning methods are selected, including PointNet++ (PN++) [50], Dynamic Graph CNN (DGCNN) [58], Pointwise CNN (PwCNN) [34], PointCNN (PCNN) [41], Laplacian Surface Network (Laplace) [40], PCP-Net (PCPN) [31] and Point Convolutional Neural Networks by Extension Operators (ExtOp) [13]. Of these methods, Laplace operates on triangle mesh input and the rest on point cloud input.…”

Section: Discussionmentioning

confidence: 99%

“…Point Convolutional Neural Networks by Extension Operators [13] is a fundamentally different way to process point clouds through mapping point cloud functions to volumetric functions and vice versa through extension and re-striction operators. A similar volumetric approach has been proposed in Pointwise Convolutional Neural Networks [34] for learning pointwise features from point clouds.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

ABC: A Big CAD Model Dataset for Geometric Deep Learning

Koch¹,

Matveev

Jiang

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

341

215

View full text Add to dashboard Cite

We introduce ABC-Dataset, a collection of one million Computer-Aided Design (CAD) models for research of geometric deep learning methods and applications. Each model is a collection of explicitly parametrized curves and surfaces, providing ground truth for differential quantities, patch segmentation, geometric feature detection, and shape reconstruction. Sampling the parametric descriptions of surfaces and curves allows generating data in different formats and resolutions, enabling fair comparisons for a wide range of geometric learning algorithms. As a use case for our dataset, we perform a large-scale benchmark for estimation of surface normals, comparing existing data driven methods and evaluating their performance against both the ground truth and traditional normal estimation methods.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

ABC: A Big CAD Model Dataset for Geometric Deep Learning

Koch¹,

Matveev

Jiang

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

341

215

View full text Add to dashboard Cite

show abstract

“…1, we compare our method with several state-ofthe-art methods in the shape classification results on both ModelNet10 and ModelNet40 datasets. Our model achieves better accuracy among the point cloud based methods (with 1024 points), such as PointNet [26], PointNet++ [28] (5K points + normals), Kd-Net (depth 15) [13], Pointwise CNN [9], KCNet [32], PointGrid [16], PCNN [3], and PointCNN [18]. Our model is slightly better than Point2Sequence [19] on ModelNet10 and shows comparable performance on ModelNet40.…”

Section: Point Cloud Classificationmentioning

confidence: 93%

A-CNN: Annularly Convolutional Neural Networks on Point Clouds

Komarichev

Zhong

Hua

2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)

265

139

View full text Add to dashboard Cite

Analyzing the geometric and semantic properties of 3D point clouds through the deep networks is still challenging due to the irregularity and sparsity of samplings of their geometric structures. This paper presents a new method to define and compute convolution directly on 3D point clouds by the proposed annular convolution. This new convolution operator can better capture the local neighborhood geometry of each point by specifying the (regular and dilated) ring-shaped structures and directions in the computation. It can adapt to the geometric variability and scalability at the signal processing level. We apply it to the developed hierarchical neural networks for object classification, part segmentation, and semantic segmentation in large-scale scenes. The extensive experiments and comparisons demonstrate that our approach outperforms the state-of-the-art methods on a variety of standard benchmark datasets (e.g., ModelNet10, ModelNet40, ShapeNetpart, S3DIS, and ScanNet).

show abstract

“…Shifting (l 2 ) [34] Adding (Hausdorff) [34] Adding (Chamfer) [34] Dropping 100 points [42] Dropping Table 3: Black-box attacks and defenses: accuracy of targeted C&W and shifting and adding based adversarial point clouds and saliency map and points dropping based adversarial point clouds generated from PointNet on other classification networks [26,32,11] with or without defense. l 2 metric, SOR has the best performance with 81.4% accuracy.…”

Section: Network Taskmentioning

confidence: 99%

DUP-Net: Denoiser and Upsampler Network for 3D Adversarial Point Clouds Defense

Zhou

Chen

Zhang

et al. 2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

129

124

View full text Add to dashboard Cite

Neural networks are vulnerable to adversarial examples, which poses a threat to their application in security sensitive systems. We propose a Denoiser and UPsampler Network (DUP-Net) structure as defenses for 3D adversarial point cloud classification, where the two modules reconstruct surface smoothness by dropping or adding points. In this paper, statistical outlier removal (SOR) and a datadriven upsampling network are considered as denoiser and upsampler respectively. Compared with baseline defenses, DUP-Net has three advantages. First, with DUP-Net as a defense, the target model is more robust to white-box adversarial attacks. Second, the statistical outlier removal provides added robustness since it is a non-differentiable denoising operation. Third, the upsampler network can be trained on a small dataset and defends well against adversarial attacks generated from other point cloud datasets. We conduct various experiments to validate that DUP-Net is very effective as defense in practice. Our best defense eliminates 83.8% of C&W and l 2 loss based attack (point shifting), 50.0% of C&W and Hausdorff distance loss based attack (point adding) and 9.0% of saliency map based attack (point dropping) under 200 dropped points on PointNet.

show abstract

Pointwise Convolutional Neural Networks

Cited by 485 publications

References 38 publications

ABC: A Big CAD Model Dataset for Geometric Deep Learning

ABC: A Big CAD Model Dataset for Geometric Deep Learning

A-CNN: Annularly Convolutional Neural Networks on Point Clouds

DUP-Net: Denoiser and Upsampler Network for 3D Adversarial Point Clouds Defense

Contact Info

Product

Resources

About