We present 3DRegNet, a novel deep learning architecture for the registration
of 3D scans. Given a set of 3D point correspondences, we build a deep neural
network to address the following two challenges: (i) classification of the
point correspondences into inliers/outliers, and (ii) regression of the motion
parameters that align the scans into a common reference frame. With regard to
regression, we present two alternative approaches: (i) a Deep Neural Network
(DNN) registration and (ii) a Procrustes approach using SVD to estimate the
transformation. Our correspondence-based approach achieves a higher speedup
compared to competing baselines. We further propose the use of a refinement
network, which consists of a smaller 3DRegNet as a refinement to improve the
accuracy of the registration. Extensive experiments on two challenging datasets
demonstrate that we outperform other methods and achieve state-of-the-art
results. The code is available.
Description
3DRegNet: A Deep Neural Network for 3D Point Registration
%0 Generic
%1 pais20193dregnet
%A Pais, G. Dias
%A Ramalingam, Srikumar
%A Govindu, Venu Madhav
%A Nascimento, Jacinto C.
%A Chellappa, Rama
%A Miraldo, Pedro
%D 2019
%K machinelearn
%T 3DRegNet: A Deep Neural Network for 3D Point Registration
%U http://arxiv.org/abs/1904.01701
%X We present 3DRegNet, a novel deep learning architecture for the registration
of 3D scans. Given a set of 3D point correspondences, we build a deep neural
network to address the following two challenges: (i) classification of the
point correspondences into inliers/outliers, and (ii) regression of the motion
parameters that align the scans into a common reference frame. With regard to
regression, we present two alternative approaches: (i) a Deep Neural Network
(DNN) registration and (ii) a Procrustes approach using SVD to estimate the
transformation. Our correspondence-based approach achieves a higher speedup
compared to competing baselines. We further propose the use of a refinement
network, which consists of a smaller 3DRegNet as a refinement to improve the
accuracy of the registration. Extensive experiments on two challenging datasets
demonstrate that we outperform other methods and achieve state-of-the-art
results. The code is available.
@misc{pais20193dregnet,
abstract = {We present 3DRegNet, a novel deep learning architecture for the registration
of 3D scans. Given a set of 3D point correspondences, we build a deep neural
network to address the following two challenges: (i) classification of the
point correspondences into inliers/outliers, and (ii) regression of the motion
parameters that align the scans into a common reference frame. With regard to
regression, we present two alternative approaches: (i) a Deep Neural Network
(DNN) registration and (ii) a Procrustes approach using SVD to estimate the
transformation. Our correspondence-based approach achieves a higher speedup
compared to competing baselines. We further propose the use of a refinement
network, which consists of a smaller 3DRegNet as a refinement to improve the
accuracy of the registration. Extensive experiments on two challenging datasets
demonstrate that we outperform other methods and achieve state-of-the-art
results. The code is available.},
added-at = {2023-09-05T16:18:13.000+0200},
author = {Pais, G. Dias and Ramalingam, Srikumar and Govindu, Venu Madhav and Nascimento, Jacinto C. and Chellappa, Rama and Miraldo, Pedro},
biburl = {https://www.bibsonomy.org/bibtex/2374a3846a0390d5df6c2a74416a8d012/cmcneile},
description = {3DRegNet: A Deep Neural Network for 3D Point Registration},
interhash = {d7d490b2b75f8a222f7a70de0b8471ff},
intrahash = {374a3846a0390d5df6c2a74416a8d012},
keywords = {machinelearn},
note = {cite arxiv:1904.01701Comment: 15 pages, 8 figures, 6 tables},
timestamp = {2023-09-05T16:18:13.000+0200},
title = {3DRegNet: A Deep Neural Network for 3D Point Registration},
url = {http://arxiv.org/abs/1904.01701},
year = 2019
}