Point Cloud Registration

184 papers with code • 22 benchmarks • 10 datasets

Point Cloud Registration is a fundamental problem in 3D computer vision and photogrammetry. Given several sets of points in different coordinate systems, the aim of registration is to find the transformation that best aligns all of them into a common coordinate system. Point Cloud Registration plays a significant role in many vision applications such as 3D model reconstruction, cultural heritage management, landslide monitoring and solar energy analysis.

Source: Iterative Global Similarity Points : A robust coarse-to-fine integration solution for pairwise 3D point cloud registration

Benchmarks

Add a Result

These leaderboards are used to track progress in Point Cloud Registration

Dataset	Best Model	Compare
ETH (trained on 3DMatch)	GeDi	See all
3DMatch Benchmark	IMFNet	See all
KITTI (trained on 3DMatch)	GeDi	See all
3DMatch (at least 30% overlapped - FCGF setting)	NgeNet	See all
3DLoMatch (10-30% overlap)	GeoTransformer - P2PNet	See all
KITTI (FCGF setting)	GeoTransformer	See all
3DMatch (at least 30% overlapped - sample 5k interest points)	NgeNet	See all
FPv1	Greedy Grid Search	See all
KITTI	GeDi	See all
FP-R-E	GeDi	See all
FP-R-M	GeDi	See all
FP-R-H	GeDi	See all
FP-T-M	Exhaustive Grid Search	See all
FP-T-E	FPFH+SC2-PCR	See all
FP-T-H	GeDi	See all
FP-O-E	FPFH+SC2-PCR	See all
FP-O-M	Exhaustive Grid Search	See all
FP-O-H	FPFH+SC2-PCR	See all
3DMatch (trained on KITTI)	GeDi	See all
KITTI (Distant PCR)	GCL+KPConv	See all
nuScenes (Distant PCR)	GCL+KPConv	See all
3RScan	SGAligner	See all

Show all 22 benchmarks

Collapse benchmarks

Libraries

Use these libraries to find Point Cloud Registration models and implementations

vinits5/learning3d

8 papers

646

chrischoy/FCGF

3 papers

606

Datasets

Subtasks

Image to Point Cloud Registration

Most implemented papers

Most implemented Social Latest No code

Open3D: A Modern Library for 3D Data Processing

isl-org/Open3D • • 30 Jan 2018

The Open3D frontend exposes a set of carefully selected data structures and algorithms in both C++ and Python.

Paper
Code

SpineNet: Learning Scale-Permuted Backbone for Recognition and Localization

tensorflow/models • • CVPR 2020

We propose SpineNet, a backbone with scale-permuted intermediate features and cross-scale connections that is learned on an object detection task by Neural Architecture Search.

Paper
Code

PointNetLK: Robust & Efficient Point Cloud Registration using PointNet

hmgoforth/PointNetLK • • CVPR 2019

To date, the successful application of PointNet to point cloud registration has remained elusive.

Paper
Code

PCRNet: Point Cloud Registration Network using PointNet Encoding

vinits5/learning3d • • 21 Aug 2019

PointNet has recently emerged as a popular representation for unstructured point cloud data, allowing application of deep learning to tasks such as object detection, segmentation and shape completion.

Paper
Code

TEASER: Fast and Certifiable Point Cloud Registration

MIT-SPARK/TEASER-plusplus • 21 Jan 2020

We propose the first fast and certifiable algorithm for the registration of two sets of 3D points in the presence of large amounts of outlier correspondences.

Paper
Code

RPM-Net: Robust Point Matching using Learned Features

yewzijian/RPMNet • • CVPR 2020

The hard assignments of closest point correspondences based on spatial distances are sensitive to the initial rigid transformation and noisy/outlier points, which often cause ICP to converge to wrong local minima.

Paper
Code

PREDATOR: Registration of 3D Point Clouds with Low Overlap

ShengyuH/OverlapPredator • • CVPR 2021

We introduce PREDATOR, a model for pairwise point-cloud registration with deep attention to the overlap region.

Paper
Code

Deep Closest Point: Learning Representations for Point Cloud Registration

WangYueFt/dcp • • ICCV 2019

To address local optima and other difficulties in the ICP pipeline, we propose a learning-based method, titled Deep Closest Point (DCP), inspired by recent techniques in computer vision and natural language processing.

Paper
Code