SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration

Ao, Sheng; Hu, Qingyong; Yang, Bo; Markham, Andrew; Guo, Yulan

doi:10.48550/arxiv.2011.12149

Cited by 4 publications

(6 citation statements)

References 60 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3DSmoothNet first uses local reference frame to orient patches, then uses a 3D convolutional neural network on the patches with a smooth density to compute features. Following the success of 3DSmoothNet, MultiView [25], DIP [29] and SpinNet [2] also improved upon the generalization capabilities from the 3DMatch dataset [51] to the ETH dataset [30]. But even if these methods are rotation invariant, patch-based methods are very slow.…”

Section: End-to-end Methodsmentioning

confidence: 99%

“…For ETH, the evaluation protocol that we have chosen is different from previous papers. We prefer using the more rigorous protocol described by Fontana et al [16]: we compare our method on the 8 scenes of the dataset and not just the 4 included in previous published works like [4,14,2,29,19].…”

Section: Eth Dataset [30] (Target Dataset)mentioning

confidence: 99%

“…As in previous works [19,4,14,29], we will assign τ 1 = 0.1 m and τ 2 = 0.05 by default. As [19,4,2,29,25], we will perform a symmetric test to filter the matches before evaluation. This metric allows evaluating feature matching.…”

Section: Metricsmentioning

confidence: 99%

See 2 more Smart Citations

3D Point Cloud Registration with Multi-Scale Architecture and Unsupervised Transfer Learning

Horache¹,

Deschaud²,

Goulette³

2021

Preprint

View full text Add to dashboard Cite

We present MS-SVConv, a fast multi-scale deep neural network that outputs features from point clouds for 3D registration between two scenes. We compute features using a 3D sparse voxel convolutional network on a point cloud at different scales and then fuse the features through fullyconnected layers. With supervised learning, we show significant improvements compared to state-of-the-art methods on the competitive and well-known 3DMatch benchmark. We also achieve a better generalization through different source and target datasets, with very fast computation. Finally, we present a strategy to fine-tune MS-SVConv on unknown datasets in a self-supervised way, which leads to state-of-the-art results on ETH and TUM datasets. The code is publicly available at: https://github.com/ humanpose1/MS-SVConv.

show abstract

Section: End-to-end Methodsmentioning

confidence: 99%

Section: Eth Dataset [30] (Target Dataset)mentioning

confidence: 99%

See 1 more Smart Citation

3D Point Cloud Registration with Multi-Scale Architecture and Unsupervised Transfer Learning

Horache¹,

Deschaud²,

Goulette³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…These descriptors are usually designed for TLS or MLS point clouds and their generalization potential to ALS point clouds is uncertain. Therefore we tested multiple 3D descriptors, namely FPFH (Rusu et al, 2009), SHOT (Tombari et al, 2010b), USC (Tombari et al, 2010a), SpinNet (Ao et al, 2021), FCGF (Choy et al, 2019) and LCD (Pham et al, 2019). While not being fully detailed here, we provide some evaluation of their suitability with respect to our ALS experimental data in Appendix 1.…”

Section: Point Description and Matchingmentioning

confidence: 99%

LiDAR Point--to--point Correspondences for Rigorous Registration of Kinematic Scanning in Dynamic Networks

Brun¹,

Cucci²,

Skaloud³

2022

Preprint

View full text Add to dashboard Cite

With the objective of improving the registration of LiDAR point clouds produced by kinematic scanning systems, we propose a novel trajectory adjustment procedure that leverages on the automated extraction of selected reliable 3D point-to-point correspondences between overlapping point clouds and their joint integration (adjustment) together with all raw inertial and GNSS observations. This is performed in a tightly coupled fashion using a Dynamic Network approach that results in an optimally compensated trajectory through modeling of errors at the sensor, rather than the trajectory, level. The 3D correspondences are formulated as static conditions within this network and the registered point cloud is generated with higher accuracy utilizing the corrected trajectory and possibly other parameters determined within the adjustment. We first describe the method for selecting correspondences and how they are inserted into the Dynamic Network as new observation models. We then describe the experiments conducted to evaluate the performance of the proposed framework in practical airborne laser scanning scenarios with low-cost MEMS inertial sensors. In the conducted experiments, the method proposed to establish 3D correspondences is effective in determining point-to-point matches across a wide range of geometries such as trees, buildings and cars. Our results demonstrate that the method improves the point cloud registration accuracy, that is otherwise strongly affected by errors in the determined platform attitude or position (in nominal and emulated GNSS outage conditions), and possibly determine unknown boresight angles using only a fraction of the total number of 3D correspondences that are established.

show abstract

“…To extract better geometrical features, Graphite [11] utilizes graph neural networks for local patch description. SpinNet [30] utilizes LRF for patch alignment and 3D cylindrical convolution layers for feature extraction, achieving the best generalization ability to unseen datasets. However, patch-based methods usually suffer from low computational efficiency, as typically shared activations of adjacent patches are not reused.…”

Section: Related Workmentioning

confidence: 99%

CoFiNet: Reliable Coarse-to-fine Correspondences for Robust Point Cloud Registration

Yu¹,

Li²,

Saleh³

et al. 2021

Preprint

View full text Add to dashboard Cite

We study the problem of extracting correspondences between a pair of point clouds for registration. For correspondence retrieval, existing works benefit from matching sparse keypoints detected from dense points but usually struggle to guarantee their repeatability. To address this issue, we present CoFiNet -Coarse-to-Fine Network which extracts hierarchical correspondences from coarse to fine without keypoint detection. On a coarse scale and guided by a weighting scheme, our model firstly learns to match down-sampled nodes whose vicinity points share more overlap, which significantly shrinks the search space of a consecutive stage. On a finer scale, node proposals are consecutively expanded to patches that consist of groups of points together with associated descriptors. Point correspondences are then refined from the overlap areas of corresponding patches, by a density-adaptive matching module capable to deal with varying point density. Extensive evaluation of CoFiNet on both indoor and outdoor standard benchmarks shows our superiority over existing methods. Especially on 3DLoMatch where point clouds share less overlap, CoFiNet significantly outperforms state-of-the-art approaches by at least 5% on Registration Recall, with at most two-third of their parameters. [Code]

show abstract

SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration

Cited by 4 publications

References 60 publications

3D Point Cloud Registration with Multi-Scale Architecture and Unsupervised Transfer Learning

3D Point Cloud Registration with Multi-Scale Architecture and Unsupervised Transfer Learning

LiDAR Point--to--point Correspondences for Rigorous Registration of Kinematic Scanning in Dynamic Networks

CoFiNet: Reliable Coarse-to-fine Correspondences for Robust Point Cloud Registration

Contact Info

Product

Resources

About