Deep Global Registration

By Christopher Bongsoo Choy, Wei Dong, Vladlen Koltun. Computer Vision and Pattern Recognition, 2020

605 citations recorded by the source corpus.

We present Deep Global Registration, a differentiable framework for pairwise registration of real-world 3D scans. Deep global registration is based on three modules: a 6-dimensional convolutional network for correspondence confidence prediction, a differentiable Weighted Procrustes algorithm for closed-form pose estimation, and a robust gradient-based SE(3) optimizer for pose refinement. Experiments demonstrate that our approach outperforms state-of-the-art methods, both learning-based and classical, on real-world data.

Subfields: Aerospace Engineering, Computational Mechanics, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/eaf4bf75c273b53e03f8743905c1760dfc94b65c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

1. Xiaoshui Huang
2. Guofeng Mei
3. Rana Abbas
4. Jian Zhang
5. Lei Li
6. Zheng Qin
7. Konrad Schindler
8. Lei Zhou

In Computational Mechanics

1. Rana Abbas
2. Jian Zhang
3. Xiaoshui Huang
4. Guofeng Mei
5. Lei Li
6. Zheng Qin
7. Konrad Schindler
8. Lei Zhou

In Computer Vision and Pattern Recognition

1. Zixin Luo
2. Andrew Markham
3. Yuxing Peng
4. Yulan Guo
5. Bo Yang
6. Lei Zhou
7. Qingyong Hu
8. Juyong Zhang

Citations in the corpus

• SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration (2020, background): “Accurate matching of partial 3D surfaces is critical for point cloud registration [17, 38, 6, 21, 29], segmentation [58, 28, 27], and recognition [24, 12, 47].”
• PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency (2021, methodology): “5, previous works [16, 48] mainly adopt node-wise losses, which supervise each correspondence individually.”
• Geometric Transformer for Fast and Robust Point Cloud Registration (2022, methodology): “As traiditional robust estimators such as RANSAC suffer from slow convergence and instability in case of high outlier ratio, deep robust estimators [3, 6, 20] have been proposed as the alternatives for them.”
• A comprehensive survey on point cloud registration (2021, methodology): “For example, DGR [16] obtains 3cm registration accuracy with ICP refinement while the accuracy drops to 22cm without ICP in the KITTI dataset.”
• Fast and Robust Iterative Closest Point (2020, methodology): “…ICP-based methods, we also compare with other methods including CPD [25] and GMM-Reg [26] based on statistical frameworks, Teaser++ [34] which uses truncated least squares optimization, as well as DCP [35] and DGR [36] based on deep learning, using their open-source implementations 3 4 5 6 7 .”
• PREDATOR: Registration of 3D Point Clouds with Low Overlap (2020, methodology): “We failed to train PRNet [41] due to random crashes of the original code (also observed in [6]).”