PCPNet Learning Local Shape Properties from Raw Point Clouds

By Paul Guerrero, Yanir Kleiman, M. Ovsjanikov, Niloy J. Mitra. 2017

426 citations recorded by the source corpus.

In this paper, we propose PCPNET, a deep‐learning based approach for estimating local 3D shape properties in point clouds. In contrast to the majority of prior techniques that concentrate on global or mid‐level attributes, e.g., for shape classification or semantic labeling, we suggest a patch‐based learning method, in which a series of local patches at multiple scales around each point is encoded in a structured manner. Our approach is especially well‐adapted for estimating local shape properties such as normals (both unoriented and oriented) and curvature from raw point clouds in the presence of strong noise and multi‐scale features. Our main contributions include both a novel multi‐scale variant of the recently proposed PointNet architecture with emphasis on local shape information, and a series of novel applications in which we demonstrate how learning from training data arising from well‐structured triangle meshes, and applying the trained model to noisy point clouds can produce superior results compared to specialized state‐of‐the‐art techniques. Finally, we demonstrate the utility of our approach in the context of shape reconstruction, by showing how it can be used to extract normal orientation information from point clouds.

Subfields: Computational Mechanics, Computer Graphics and Computer-Aided Design, 3D Vision & Pose

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Mechanics

In Computer Graphics and Computer-Aided Design

In 3D Vision & Pose

Citations in the corpus

Fast Graph Representation Learning with PyTorch Geometric (2019, methodology): “In addition, we provide embedded datasets like MNIST superpixels (Monti et al., 2017), FAUST (Bogo et al., 2014), ModelNet10/40 (Wu et al., 2015), ShapeNet (Chang et al., 2015), COMA (Ranjan et al., 2018), and the PCPNet dataset from Guerrero et al. (2018).”
DensePoint: Learning Densely Contextual Representation for Efficient Point Cloud Processing (2019, background): “Accordingly, some works [34, 5, 39] partition point cloud into local subsets and learn on them based on PointNet.”
Patch-Based Progressive 3D Point Set Upsampling (2018, methodology): “It has been successfully applied in point set segmentation [10, 40], generation [1,13,56], consolidation [14,45,58], deformation [57], completion [15, 60] and upsampling [58, 59, 61].”
Deep Hough Voting for 3D Object Detection in Point Clouds (2019, methodology): “While our method is not restricted to any point cloud network, we adopt PointNet++ [36] as our backbone due to its simplicity and demonstrated success on tasks ranging from normal estimation [10], semantic segmentation [21] to 3D object localization [34].”
Differentiable surface splatting for point-based geometry processing (2019, background): “Based on PointNet [Qi et al. 2017a], PCPNET [Guerrero et al. 2018] and PointCleanNet [Rakotosaona et al. 2019] 2014] and NMR [Kato et al. 2018] do not propagate gradients to depth; Paparazzi [Liu et al. 2018] has limitation in updating the vertex positions in directions orthogonal their face…”
Shape As Points: A Differentiable Poisson Solver (2021, positive): “Most reconstruction approaches directly reconstruct a meshed surface geometry, though some works [3,4,21,33] ﬁrst predict point set normals to subsequently reconstruct the geometry via PSR [30,31].”
Relation-Shape Convolutional Neural Network for Point Cloud Analysis (2019, neutral): “Local structure exploitation with PointNet is also investigated in [4, 30].”
PointASNL: Robust Point Clouds Processing Using Nonlocal Neural Networks With Adaptive Sampling (2020, methodology): “Various data-driven methods [9, 26] are proposed for outlier removal, which first discard some outliers and then projects noisy points to clean surfaces.”
ABC: A Big CAD Model Dataset for Geometric Deep Learning (2018, methodology): “PCPNet: Learning Local Shape Properties from Raw Point Clouds [27].”
PU-Net: Point Cloud Upsampling Network (2018, background): “[10] developed a network to estimate the local shape properties in point clouds, including normal and curvature.”
Monte Carlo convolution for learning on non-uniformly sampled point clouds (2018, background): “PCP Net [Guerrero et al. 2018] allowed the inference of local properties like curvature or normals but was also not convolutional.”
PointCleanNet: Learning to Denoise and Remove Outliers from Dense Point Clouds (2019, methodology): “In settings such as ours, local meth-ods such as PCPNet [GKOM18] perform better.”
P2P-NET: Bidirectional Point Displacement Net for Shape Transform (methodology): “Recently, deep neural networks have been designed to learn global and multi-scale point set features for shape classification and segmentation [Qi et al. 2017a,b], as well as geometry (e.g., normal or curvature) estimation [Guerrero et al. 2017].”
Multi-directional geodesic neural networks via equivariant convolution (2018, methodology): “et [Qi et al. 2017], PointNet++ [Qi et al. 2017]and several extensions such as PointCNN [Li et al. 2018] and Dynamic Graph CNN [Wang et al. 2018]for shape classification and segmentation, and PCPNet [Guerrero et al. 2018] for normal and curvature estimation, among others. Despite their efficiency an”
Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution (2018, background): “ouds, which demonstrate its advantages in real-world data analysis. P2P-Net[31] applies bidirectional networks and extend PointNet++ to learn geometric transformations between two point clouds. PCPNet[6] and PointProNets[21] are designed to learn normals and curvatures on raw point clouds and ﬁt it ”
EC-Net: an Edge-aware Point set Consolidation Network (2018, background): “Particularly, the emergence of neural networks and their startling performance provide a new means for 3D reconstruction from point sets by data-driven learning [6,7,8].”