The repository mainly focuses on collecting recent works on point cloud registration, for sharing and better learning. Since many other tasks, like image matching and point cloud processing, are highly related to point cloud registration, these papers will be taken into account as well. In order to have a clear look at these fields, We simply classified the articles.This repository will continue to be updated.
labels of task: [cla.] for classfication | [reg.] for registration | [def] for shape deformation | [gen.] for shape generation | [nor.] for normal estimation | [shot.] for few/one/zero-shot learning | [recon.] for reconstruction.
Statistics: 🔥 code is available & stars >= 100 | ⭐ citation >= 50
what
-
Point-based methods
-
Voxel-based methods
-
Generative model methods
[CVPR 2020] PointGMM: a Neural GMM Network for Point Clouds. [Paper][Code] [reg. gen.]
[ECCV 2018] Hgmr: Hierarchical gaussian mixtures for adaptive 3d registration. [Paper] [reg.]
-
Lattice method
-
Tangent convolution method
[Com. ACM 1981] Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography.(RANSAC) [Paper]
- Keypoint-based methods (Sparse)
[CoRR 2020] D2D: Learning to find good correspondences for image matching and manipulation [Paper]
- Dense methods
[CVPR 2017] DeMoN: Depth and Motion Network for Learning Monocular Stereo. [Paper]
[CVPR 2017] DSAC-Differentiable RANSAC for Camera Localization. [Paper]
[CVPR 2018] Learning to Find Good Correspondences. [Paper]
- Point-based methods
[CVPR 2019] PointConv: Deep Convolutional Networks on 3D Point Clouds.[Paper][cla. seg.]
[CVPR 2018] PointNet
- Voxel-based methods
SparseConvNet[Code]
[CVPR 2018] SyncSpecCNN: Synchronized Spectral CNN for 3D Shape Segmentation. [Paper] [seg.]
- Generative model methods
[CVPR 2019] Nesti-net: Normal estimation for unstructured 3d pointclouds using convolutional neural networks. [Papers] [nor.]
[Graphic 2019] SDM-NET: deep generative network for structured deformable mesh. [Paper][def.]
[TOG 2019] SDM-NET: Deep generative network for structured deformable mesh. [Paper] ][gen.]
[R&A Letters 2018] 3dmfv: Three-dimensional point cloud classification in real-time using convolutional neural networks. [Paper] [cla.]
- Spherical CNN
- Else
[ICLR 2020] Learning to Group: A Bottom-Up Framework for 3D Part Discovery in Unseen Categories. [Paper]
[ACM Graphic 2018] Deep part induction from articulated object pairs. [Paper][shot.]
[IJCV 2020] Image Matching from Handcrafted to Deep Features: A Survey. [Paper]
[SPM 2017] Geometric Deep Learning. [Paper]
[ICCV 2019] Deep Mesh Reconstruction From Single RGB Images via Topology Modification Networks.[Paper]
- Synthetic dataset
ShapeNet: Large Scale Synthetic Objects
ModelNet
PartNet:Fine-grained; Instance-level; Hierarchical.
SceneNet: Large Scale Synthetic Scenes; 3D meshes
- Real world dataset
3DScan
ScanNet: Large-scale; RGBD scans; 3D camera poses; Instance-level segmentation.
KITTI: LiDAR data; labeled by 3D b.boxes.
Semantic KITTI: LiDAR data; labeled per point.
Waymo Open Dataset: LiDAR data; labeled by 3D b.boxes.