3D Feature Matching
10 papers with code • 1 benchmarks • 4 datasets
Image: Choy et al
Benchmarks
These leaderboards are used to track progress in 3D Feature Matching
| Trend | Dataset | Best Model | Paper | Code | Compare |
|---|---|---|---|---|---|
|
3DMatch
DeformingThings4D
4DMatch
CPNet
Most implemented papers
3D Point Capsule Networks
yongheng1991/3D-point-capsule-networks
•
•
CVPR 2019
In this paper, we propose 3D point-capsule networks, an auto-encoder designed to process sparse 3D point clouds while preserving spatial arrangements of the input data.
PRIN/SPRIN: On Extracting Point-wise Rotation Invariant Features
qq456cvb/SPRIN
•
•
Spherical Voxel Convolution and Point Re-sampling are proposed to extract rotation invariant features for each point.
Pointwise Rotation-Invariant Network with Adaptive Sampling and 3D Spherical Voxel Convolution
qq456cvb/PRIN
•
•
Point cloud analysis without pose priors is very challenging in real applications, as the orientations of point clouds are often unknown.
Fully Convolutional Geometric Features
chrischoy/FCGF
•
•
International Conference on Computer vision 2019
Extracting geometric features from 3D scans or point clouds is the first step in applications such as registration, reconstruction, and tracking.
Human Correspondence Consensus for 3D Object Semantic Understanding
yokinglou/CorresPondenceNet
•
•
ECCV 2020
Semantic understanding of 3D objects is crucial in many applications such as object manipulation.
Towards Robust Visual Information Extraction in Real World: New Dataset and Novel Solution
Visual information extraction (VIE) has attracted considerable attention recently owing to its various advanced applications such as document understanding, automatic marking and intelligent education.
Lepard: Learning partial point cloud matching in rigid and deformable scenes
We present Lepard, a Learning based approach for partial point cloud matching in rigid and deformable scenes.
Points to Patches: Enabling the Use of Self-Attention for 3D Shape Recognition
While the Transformer architecture has become ubiquitous in the machine learning field, its adaptation to 3D shape recognition is non-trivial.
Improving Feature-based Visual Localization by Geometry-Aided Matching
We apply GAM to a new hierarchical visual localization pipeline and show that GAM can effectively improve the robustness and accuracy of localization.
EP2P-Loc: End-to-End 3D Point to 2D Pixel Localization for Large-Scale Visual Localization
minnjung/ep2p-loc
•
•
ICCV 2023
Visual localization is the task of estimating a 6-DoF camera pose of a query image within a provided 3D reference map.
