KPConv: Flexible and Deformable Convolution for Point Clouds (2019-04-18T00:00:00.000000Z)

TL;DR

KPConv is a new design of point convolution, i.e. that operates on point clouds without any intermediate representation, that outperform state-of-the-art classification and segmentation approaches on several datasets.

Abstract

We present Kernel Point Convolution (KPConv), a new design of point convolution, i.e. that operates on point clouds without any intermediate representation. The convolution weights of KPConv are located in Euclidean space by kernel points, and applied to the input points close to them. Its capacity to use any number of kernel points gives KPConv more flexibility than fixed grid convolutions. Furthermore, these locations are continuous in space and can be learned by the network. Therefore, KPConv can be extended to deformable convolutions that learn to adapt kernel points to local geometry. Thanks to a regular subsampling strategy, KPConv is also efficient and robust to varying densities. Whether they use deformable KPConv for complex tasks, or rigid KPconv for simpler tasks, our networks outperform state-of-the-art classification and segmentation approaches on several datasets. We also offer ablation studies and visualizations to provide understanding of what has been learned by KPConv and to validate the descriptive power of deformable KPConv.

Authors

C. Qi

15 papers

L. Guibas

28 papers

Jean-Emmanuel Deschaud

3 papers

TL;DR

Abstract

Authors

References55 items

Point2Sequence: Learning the Shape Representation of 3D Point Clouds with an Attention-based Sequence to Sequence Network

3D Recurrent Neural Networks with Context Fusion for Point Cloud Semantic Segmentation

Efficient Convolutions for Real-Time Semantic Segmentation of 3D Point Clouds

Fully-Convolutional Point Networks for Large-Scale Point Clouds

Semantic Classification of 3D Point Clouds with Multiscale Spherical Neighborhoods

3DmFV: Three-Dimensional Point Cloud Classification in Real-Time Using Convolutional Neural Networks

Monte Carlo convolution for learning on non-uniformly sampled point clouds

Tangent Convolutions for Dense Prediction in 3D

Deep Parametric Continuous Convolutional Neural Networks

Classification of Point Cloud Scenes with Multiscale Voxel Deep Network

SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters

Point convolutional neural networks by extension operators

Local Spectral Graph Convolution for Point Set Feature Learning

SO-Net: Self-Organizing Network for Point Cloud Analysis

Flex-Convolution - Million-Scale Point-Cloud Learning Beyond Grid-Worlds

SPLATNet: Sparse Lattice Networks for Point Cloud Processing

Recurrent Slice Networks for 3D Segmentation of Point Clouds

Dynamic Graph CNN for Learning on Point Clouds

PointCNN: Convolution On X-Transformed Points

Mining Point Cloud Local Structures by Kernel Correlation and Graph Pooling

Pointwise Convolutional Neural Networks

Paris-Lille-3D: A large and high-quality ground-truth urban point cloud dataset for automatic segmentation and classification

3D Semantic Segmentation with Submanifold Sparse Convolutional Networks

Large-Scale Point Cloud Semantic Segmentation with Superpoint Graphs

SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation

SEGCloud: Semantic Segmentation of 3D Point Clouds

FeaStNet: Feature-Steered Graph Convolutions for 3D Shape Analysis

PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space

Deep Projective 3D Semantic Segmentation

Semantic3D.net: A new Large-scale Point Cloud Classification Benchmark

Dynamic Edge-Conditioned Filters in Convolutional Neural Networks on Graphs

Escape from Cells: Deep Kd-Networks for the Recognition of 3D Point Cloud Models

Deformable Convolutional Networks

Deep Sets

ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes

Understanding the Effective Receptive Field in Deep Convolutional Neural Networks

PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation

SyncSpecCNN: Synchronized Spectral CNN for 3D Shape Segmentation

Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs

Geometric Deep Learning: Going beyond Euclidean data

OctNet: Learning Deep 3D Representations at High Resolutions

A scalable active framework for region annotation in 3D shape collections

Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering

3D Semantic Parsing of Large-Scale Indoor Spaces

Deep Residual Learning for Image Recognition

VoxNet: A 3D Convolutional Neural Network for real-time object recognition

Multi-view Convolutional Neural Networks for 3D Shape Recognition

Geodesic Convolutional Neural Networks on Riemannian Manifolds

3D ShapeNets: A deep representation for volumetric shapes

Clouds

Unstructured Point Cloud Semantic Labeling Using Deep Segmentation Networks

Method class inst. aero bag cap car chair ear guit knif lamp lapt moto mug pist rock skate table avg. avg

Segmentation mIoUs on ShapeNetPart. Method mIoU mRec ceil. floor wall beam col. wind. door chair table book

Semantic segmentation IoU scores on S3DIS k-fold. Additionally, we give the mean class recall, a measure that some previous works call mean class accuracy

A . Lensch . Flex - convolution

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names