PREDATOR: Registration of 3D Point Clouds with Low Overlap (2020-11-25T00:00:00.000000Z)

TL;DR

The ability to focus on points that are relevant for matching greatly improves performance: PREDATOR raises the rate of successful registrations by more than 20% in the low-overlap scenario, and also sets a new state of the art for the 3DMatch benchmark with 89% registration recall.

Abstract

We introduce PREDATOR, a model for pairwise point-cloud registration with deep attention to the overlap region. Different from previous work, our model is specifically designed to handle (also) point-cloud pairs with low overlap. Its key novelty is an overlap-attention block for early information exchange between the latent encodings of the two point clouds. In this way the subsequent decoding of the latent representations into per-point features is conditioned on the respective other point cloud, and thus can predict which points are not only salient, but also lie in the overlap region between the two point clouds. The ability to focus on points that are relevant for matching greatly improves performance: PREDATOR raises the rate of successful registrations by more than 20% in the low-overlap scenario, and also sets a new state of the art for the 3DMatch benchmark with 89% registration recall. [Code release]

Authors

K. Schindler

19 papers

Shengyu Huang

2 papers

Zan Gojcic

3 papers

TL;DR

Abstract

Authors

References57 items

RSKDD-Net: Random Sample-based Keypoint Detector and Descriptor

D2D: Learning to find good correspondences for image matching and manipulation

Deep Global Registration

RPM-Net: Robust Point Matching Using Learned Features

D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features

Circle Loss: A Unified Perspective of Pair Similarity Optimization

Learning Multiview 3D Point Cloud Registration

Extreme Relative Pose Network Under Hybrid Representations

SuperGlue: Learning Feature Matching With Graph Neural Networks

PRNet: Self-Supervised Learning for Partial-to-Partial Registration

Fully Convolutional Geometric Features

R2D2: Repeatable and Reliable Detector and Descriptor

Deep Closest Point: Learning Representations for Point Cloud Registration

4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks

KPConv: Flexible and Deformable Convolution for Point Clouds

3DRegNet: A Deep Neural Network for 3D Point Registration

PointNetLK: Robust & Efficient Point Cloud Registration Using PointNet

Extreme Relative Pose Estimation for RGB-D Scans via Scene Completion

The Perfect Match: 3D Point Cloud Matching With Smoothed Densities

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

3DFeat-Net: Weakly Supervised Local 3D Features for Point Cloud Registration

LEARNED COMPACT LOCAL FEATURE DESCRIPTOR FOR TLS-BASED GEODETIC MONITORING OF NATURAL OUTDOOR SCENES

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

Dynamic Graph CNN for Learning on Point Clouds

SuperPoint: Self-Supervised Interest Point Detection and Description

Learning Compact Geometric Features

Attention is All you Need

Neural Message Passing for Quantum Chemistry

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

Semi-Supervised Classification with Graph Convolutional Networks

Structured Global Registration of RGB-D Scans in Indoor Environments

Instance Normalization: The Missing Ingredient for Fast Stylization

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

Learning to Navigate the Energy Landscape

Robust reconstruction of indoor scenes

Fully convolutional networks for semantic segmentation

Performance Evaluation of 3D Local Feature Descriptors

Unsupervised feature learning for 3D scene labeling

3D ShapeNets: A deep representation for volumetric shapes

SUN3D: A Database of Big Spaces Reconstructed Using SfM and Object Labels

Scene Coordinate Regression Forests for Camera Relocalization in RGB-D Images

Are we ready for autonomous driving? The KITTI vision benchmark suite

Unique shape context for 3d data description

Unique Signatures of Histograms for Local Surface Description

Rectified Linear Units Improve Restricted Boltzmann Machines

Curriculum learning

Fast Point Feature Histograms (FPFH) for 3D registration

Aligning point cloud views using persistent feature histograms

Using Spin Images for Efficient Object Recognition in Cluttered 3D Scenes

A volumetric method for building complex models from range images

Method for registration of 3-D shapes

A Method for Registration of 3-D Shapes

Least-Squares Fitting of Two 3-D Point Sets

An Iterative Image Registration Technique with an Application to Stereo Vision

A Relationship Between Arbitrary Positive Matrices and Doubly Stochastic Matrices

Rectifier Nonlinearities Improve Neural Network Acoustic Models

Relative Rotation Error ( • ) ↓

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names