Cross-Instance Gaussian Splatting Registration via Geometry-Aware Feature-Guided Alignment

Roy Amoyal; Oren Freifeld; Chaim Baskin

cs.CVMar 23, 2026

Cross-Instance Gaussian Splatting Registration via Geometry-Aware Feature-Guided Alignment

Authors: Roy Amoyal, Oren Freifeld, Chaim Baskin

Organizations: Ben-Gurion University of the Negev, Israel

Abstract

We present Gaussian Splatting Alignment (GSA), a novel method for aligning two independent 3D Gaussian Splatting (3DGS) models via a similarity transformation (rotation, translation, and scale), even when they are of different objects in the same category (e.g., different cars). In contrast, existing methods can only align 3DGS models of the same object (e.g., the same car) and often must be given true scale as input, while we estimate it successfully. GSA leverages viewpoint-guided spherical map features to obtain robust correspondences and introduces a two-step optimization framework that aligns 3DGS models while keeping them fixed. First, we apply an iterative feature-guided absolute orientation solver as our coarse registration, which is robust to poor initialization (e.g., 180 degrees misalignment or a 10x scale gap). Next, we use a fine registration step that enforces multi-view feature consistency, inspired by inverse radiance-field formulations. The first step already achieves state-of-the-art performance, and the second further improves results. In the same-object case, GSA outperforms prior works, often by a large margin, even when the other methods are given the true scale. In the harder case of different objects in the same category, GSA vastly surpasses them, providing the first effective solution for category-level 3DGS registration and unlocking new applications. Project webpage: https://bgu-cs-vil.github.io/GSA-project/

Figures & tables

Panel of Figure 1 — Figure 1: (A) Cross-instance registration: GSA successfully aligns the red boat to the white boat despite extremely poor initialization, including a 180° rotation and over 5× scale difference. (B) GSA enables geometrically consistent object replacement even across datasets, e.g., replacing the white car from 3D Real Car [ 9 ] with purple/red cars from the same dataset and a police car from ShapeNet [ 3 ] .

Figure 2: The GSA pipeline. Please refer to the text for details on each algorithm.

Panel of Figure 3 — Figure 3: A 3DGS model with enhanced geometric semantic features [ 21 ] (a) input image; (b) rendered image; (c) low-resolution input feature map;(d) high-resolution rendered feature map.

Figure 4: The features from [ 21 ] capture both semantic similarity ( e.g . , wings, engines, tail) and spatial distinctions ( e.g . , left/right, front/back). Colors indicate matched feature regions across two airplane models; arrows highlight example correspondences.

Method	RRE (°) $\downarrow$	ATE $\downarrow$
FGR [ 41 ]	51.32	0.33
REGTR [ 36 ]	125.34	0.38
DReg-NeRF [ 5 ]	10.37	0.04
GaussReg [ 4 ]	18.60	0.07
(Ours) GSA: Coarse	3.13	0.01
(Ours) GSA: Coarse+Fine	0.36	0.001

Table 1: Quantitative comparison of registration accuracy for same-object alignment. RRE: Relative Rotation Error (in degrees), ATE: Absolute Translation Error. Lower values are better (

\downarrow

Method	Airplane	Bus	Boat	Car	Chair	Motorcycle
	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$
FGR [ 41 ]	90.78	94.98	86.60	89.32	91.53	82.06
REGTR [ 36 ]	90.00	77.69	90.98	96.98	94.75	86.97
GaussReg [ 4 ]	136.32	138.51	126.64	121.36	155.03	120.14
GSA: Coarse (Ours)	2.97	4.74	23.97	3.70	14.17	3.13
GSA: Coarse+Fine (Ours)	2.45	2.41	22.68	1.71	10.86	4.05

Table 2: Comparison of registration accuracy across categories. Each method is evaluated on six object categories.

Metric	Method	CO3Dv2	CO3Dv2	CO3Dv2	3D Real
		toyplane	bicycle	chair	car
Mean RRE (°) $\downarrow$	GaussReg	126.09	121.12	124.45	109.72
	GSA (Coarse)	7.45	9.19	49.90	2.45
	GSA (Coarse + Fine)	6.70	13.29	44.56	1.82
Median RRE (°) $\downarrow$	GaussReg	121.77	122.27	125.66	107.66
	GSA (Coarse)	6.44	4.60	33.12	1.93

Table 3: RRE (°) (

\downarrow

) on cross-instance alignment of real objects.

Panel of Figure 5 — Figure 5: Same-object alignment comparison. Only GSA accurately recovers the full similarity transformation, achieving near-perfect alignment, such that the source and target fully overlap.

Configuration	Airplane	Bus	Boat	Car	Chair	Motorcycle
	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$	RRE (°) $\downarrow$
GSA (only coarse) - 1 iteration	91.04	86.79	83.10	95.52	84.42	82.52
GSA (only coarse) - 3 iterations	3.05	4.30	25.17	3.27	14.54	2.81
GSA (only coarse) - Until convergence	2.97	4.06	30.21	3.26	10.50	2.89
GSA (coarse 3 iterations + fine) - 3 (similar) views	4.23	4.56	29.10	5.15	11.54	6.05
GSA (coarse 3 iterations + fine) - 3 (diverse) views	2.72	3.02	22.68	1.88	10.26	3.94

Table 4: Testing various GSA configurations on different object categories. Lower is better (

\downarrow

Appendix figures & tables11 assets

Supplementary material from the paper’s appendix.

Appendix

Panel of Figure 10 — Figure A.1 : A Comparison of alignment results on two 3DGS models of the same chair from the Objaverse dataset. Although the object has a clear and rigid structure, classical and learning-based methods struggle to achieve consistent alignment. Our GSA method, especially with coarse and fine refinement, achieves the most accurate and coherent results.

Panel of Figure 11 — Figure A.2 : Comparison of two 3DGS models of the same teddy bear, Objaverse [ 8 ] dataset. Although the object is somewhat simple and although there is a unique rigid transformation solution, existing point cloud-based registration methods still struggle while our coarse point-cloud registration achieves almost-perfect results (so in this particular case, the further improvement by our fine step is hard to notice visually).

Panel of Figure 12 — Figure A.3 : Visual qualitative results for aligning different objects within the same cateogry. In each of the three 2-row blocks we show an example form a different category (boats, airplanes, and chairs). The two rows within each 2-row block show different views. In each 2-row block we show, from lest to right: 1) a random initialization; 2) the result of GSA’s coarse step; 3) the result of GSA’s fine step. The random initialization includes arbitrary translations, rotations up to 180 degrees, and significant scale differences.

Panel of Figure 13 — Figure B.1 : Geometrically consistent object replacement using GSA. Each row shows a different camera view. The first column displays the original white car. The second to fourth columns show replacements with red, purple, and toy police cars, respectively. GSA aligns the models in rotation, translation, and symmetric scale, preserving their intrinsic geometry while adapting them to best match the target object’s position, orientation, and scale in a structurally faithful way.

Panel of Figure 14 — Figure B.2 : Synchronized novel view and feature renderings across multiple semantically aligned motorcycle instances. Each row shows either a novel rendering or the corresponding feature projection for different objects and camera views.

Panel of Figure 15 — Figure B.3 : Renderings of the unified 3D Gaussian Splatting (3DGS) model combining all gaussians from the aligned motorcycle instances, shown from two additional novel synchronized views. These renderings provide a global perspective of the unified category-level geometry.

Method	Mean RRE (deg)	Mean Scale Error (%)
GSA: Coarse	$0.362$	$0.21$
GSA: Coarse+Fine	$0.074$	$0.08$

Appendix

Table C.1 : Quantitative evaluation of rotation and scale estimation on transformed ShapeNet objects.

Panel of Figure 17 — Figure C.1 : Qualitative registration results. Top row: comparison of the random initialization, coarse, and fine alignment results. Bottom row: close-up views of the boat top, highlighting the improved accuracy of the fine stage over the coarse alignment.

Panel of Figure 18 — Figure D.1 : Augmented 3DGS models using different choices of semantic features: [ 21 ] , [ 25 ] , and [ 39 ] . For [ 25 ] or [ 39 ] , dimensionality was reduced to 3 by Principal Component Analysis.

Panel of Figure 19 — Figure D.2 : Qualitative comparison of the GSA coarse alignment step, when using different features. Top row: Unaligned models. Middle row, left: Alignment result when using DINOv2 [ 25 ] . Note the angular mistake is about 180 degrees. Middle row, right: Alignment results when using and [ 39 ] . Again the mistake is about 180 degrees, and the tilting angle is also evidently wrong. Bottom row: In contrast, when GSA uses [ 21 ] , it successfully aligns the cars.

Panel of Figure 20 — Figure E.1 : The effect of background pruning. (a) Extraneous background Gaussians create occlusions, and are inconsistent across the two models, complicating alignment. (b) Our solution removes these artifacts, improving alignment consistency.

3D Gaussian Splatting

3D Gaussian Splatting represents scenes as optimizable 3D Gaussian primitives for fast, photorealistic rendering from new viewpoints.

See all 1,019 papers

PointGauss: Point Cloud-Guided Multi-Object Segmentation for Gaussian Splatting
Aug 1, 2025Wentao Sun, Yiping Chen, John S. Zelek +13D Gaussian Splatting 3DGS Segmentation
University of Waterloo, Waterloo, ON N2L 3G1, Canada · Sun Yat-Sen University, Zhuhai 519082, China
Graph-GSReg: Leveraging 3D Scene Graphs for Gaussian Splatting Registration
Jun 29, 2026Jaewon Lee, Mangyu Kong, Euntai Kim3D Gaussian Splatting Graph Matching
Yonsei University, Seoul, Republic of Korea · Korea Institute of Science and Technology, Seoul, Republic of Korea
StructSplat: Generalizable 3D Gaussian Splatting from Uncalibrated Sparse Views
Jun 26, 2026Jia-Chen Zhao, Beiqi Chen, Xinyang Chen +23D Gaussian Splatting 3D Reconstruction
Harbin Institute of Technology (Shenzhen) · Great Bay University · Guangzhou CloudButterfly Technology Co., Ltd.

Aug 1, 2025J/K move · Enter open · S save

PointGauss: Point Cloud-Guided Multi-Object Segmentation for Gaussian Splatting

Wentao Sun, Yiping Chen, John S. Zelek, Jonathan Li

University of Waterloo, Waterloo, ON N2L 3G1, Canada · Sun Yat-Sen University, Zhuhai 519082, China

3D Gaussian Splatting 3DGS Segmentation 3D Semantic Segmentation Remote Sensing 3D Instance Segmentation

Open paper AskPDF arXiv

While 3D Gaussian Splatting (3DGS) has established new standards for high-fidelity 3D scene modeling, interpreting massive, unstructured Gaussian primitives into meaningful geospatial entities remains a critical challenge for remote sensing and urban applications. Existing 2D-to-3D distillation methods suffer from projection ambiguities, geometric inconsistencies, and computational bottlenecks when applied to complex, large-scale topographies. Furthermore, current benchmarks lack natively aligned 2D-3D ground truth tailored for splatting-based representations. To address these limitations, we propose PointGauss, a unified 3D-native framework for robust semantic parsing and instance segmentation. By treating Gaussian primitives as unstructured point sets, PointGauss leverages Point Transformer V3 (PTv3) to extract scale-invariant geometric features directly from Gaussian attributes, effectively resolving projection ambiguities. To ensure scalability in building-scale scenes, we introduce an adaptive region-of-interest cropping strategy and an instance-aware distance-constrained rasterization pipeline for pixel-level, view-consistent projection. Additionally, we present SplatSeg-360, the first rigorous cross-scale benchmark tailored for 3DGS, comprising 32 complex scenes with full 360

^\circ

coverage and over 6,300 natively aligned 2D-3D masks. Extensive experiments demonstrate that PointGauss operates in real-time and achieves state-of-the-art performance. Notably, it attains approximately 90% 3D-mIoU in large-scale building scenarios and roughly 80% 2D-mIoU in view-consistent 2D instance segmentation, outperforming baseline methods by 16%. (Code)

Abstract

3D Gaussian Splatting

Explore similar work

PointGauss: Point Cloud-Guided Multi-Object Segmentation for Gaussian Splatting

Graph-GSReg: Leveraging 3D Scene Graphs for Gaussian Splatting Registration

StructSplat: Generalizable 3D Gaussian Splatting from Uncalibrated Sparse Views