3D Point Capsule Networks (2018-12-27T00:00:00.000000Z)

TL;DR

3D capsule networks are proposed, an auto-encoder designed to process sparse 3D point clouds while preserving spatial arrangements of the input data and enables new applications such as part interpolation and replacement.

Abstract

In this paper, we propose 3D point-capsule networks, an auto-encoder designed to process sparse 3D point clouds while preserving spatial arrangements of the input data. 3D capsule networks arise as a direct consequence of our unified formulation of the common 3D auto-encoders. The dynamic routing scheme and the peculiar 2D latent space deployed by our capsule networks bring in improvements for several common point cloud-related tasks, such as object classification, object reconstruction and part segmentation as substantiated by our extensive evaluations. Moreover, it enables new applications such as part interpolation and replacement.

Authors

Federico Tombari

15 papers

Yongheng Zhao

1 papers

Tolga Birdal

3 papers

TL;DR

Abstract

Authors

References54 items

On Learning and Learned Data Representation by Capsule Networks

PointGrow: Autoregressively Learned Point Cloud Generation with Self-Attention

On Learning and Learned Representation with Dynamic Routing in Capsule Networks

Dynamic Programming

PPF-FoldNet: Unsupervised Learning of Rotation Invariant 3D Local Descriptors

Generalized Capsule Networks with Trainable Routing Procedure

Capsule Networks against Medical Imaging Data Challenges

3D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Data

Fast CapsNet for Lung Cancer Screening

Generative Adversarial Network Architectures For Image Synthesis Using Capsule Networks

CapsGAN: Using Dynamic Routing for Generative Adversarial Networks

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

Group Equivariant Capsule Networks

Tangent Convolutions for Dense Prediction in 3D

Fast dynamic routing based on weighted kernel density estimation

Deep Functional Dictionaries: Learning Consistent Semantic Structures on 3D Models from Functions

VideoCapsuleNet: A Simplified Network for Action Detection

Spherical Convolutional Neural Network for 3D Point Clouds

CapProNet: Deep Feature Learning via Orthogonal Projections onto Capsule Subspaces

Capsules for Object Segmentation

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

Indoor Scene Understanding in 2.5/3D for Autonomous Agents: A Survey

Indoor Scene Understanding in 2.5/3D: A Survey

CapsuleGAN: Generative Adversarial Capsule Network

A Papier-Mache Approach to Learning 3D Surface Generation

Matrix capsules with EM routing

An Optimization View on Dynamic Routing Between Capsules

PPFNet: Global Context Aware Local Features for Robust 3D Point Matching

Dynamic Graph CNN for Learning on Point Clouds

PU-Net: Point Cloud Upsampling Network

FoldingNet: Point Cloud Auto-Encoder via Deep Grid Deformation

VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection

Dynamic Routing Between Capsules

Learning Compact Geometric Features

A point sampling algorithm for 3D matching of irregular geometries

Learning Representations and Generative Models for 3D Point Clouds

Learning Local Shape Descriptors from Part Correspondences with Multiview Convolutional Networks

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

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

OctNet: Learning Deep 3D Representations at High Resolutions

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

Learning a Probabilistic Latent Space of Object Shapes via 3D Generative-Adversarial Modeling

3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions

ShapeNet: An Information-Rich 3D Model Repository

Point Pair Features Based Object Detection and Pose Estimation Revisited

3D ShapeNets: A deep representation for volumetric shapes

Transforming Auto-Encoders

Et al

Yongbin Sun

and F

GENERATIVE ADVERSARIAL NETS

c ○ 2000 Kluwer Academic Publishers. Manufactured in The Netherlands. The Earth Mover’s Distance as a Metric for Image Retrieval

We propose capsule networks for the realm of 3D data processing as a powerful and effective tool

Table 6. Part segmentation on ShapeNet-Part by learning on limited training data

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names