The increasing demand for augmented reality and robotics is driving the need for articulated object reconstruction with high scalability. However, existing settings for reconstructing from discrete articulation states or casual monocular videos require non-trivial axis alignment or suffer from insufficient coverage, limiting their applicability. In this paper, we introduce FreeArtGS, a novel method for reconstructing articulated objects under free-moving scenario, a new setting with a simple setup and high scalability. FreeArtGS combines free-moving part segmentation with joint estimation and end-to-end optimization, taking only a monocular RGB-D video as input. By optimizing with the priors from off-the-shelf point-tracking and feature models, the free-moving part segmentation module identifies rigid parts from relative motion under unconstrained capture. The joint estimation module calibrates the unified object-to-camera poses and recovers joint type and axis robustly from part segmentation. Finally, 3DGS-based end-to-end optimization is implemented to jointly reconstruct visual textures, geometry, and joint angles of the articulated object. We conduct experiments on two benchmarks and real-world free-moving articulated objects. Experimental results demonstrate that FreeArtGS consistently excels in reconstructing free-moving articulated objects and remains highly competitive in previous reconstruction settings, proving itself a practical and effective solution for realistic asset generation. The project page is available at: https://freeartgs.github.io/
Figures & tables
Figure 1: FreeArtGS reconstructs articulated object under the free-moving scenario, in which both joint and object pose move in an unconstrained manner.
Figure 2: Overview of FreeArtGS. It consists of three modules: (1) Part Segmentation from free-moving video; (2) Joint Estimation with estimated part transforms; (3) End-to-end Optimization for articulated Gaussian Splatting.
Joint Type
Method
Axis (deg) ↓
Position (cm) ↓
State (deg/cm) ↓
CD-w (cm) ↓
CD-m (cm) ↓
CD-s (cm) ↓
PSNR (dB) ↑
Revolute
ArticulateAnything [ 16 ]
42.00 ± 46.48
59.38 ± 62.19
-
-
-
-
-
Video2Articulation [ 24 ]
20.00 ± 28.81
16.31 ± 28.78
27.37 ± 19.06
2.29 ± 3.40
10.74 ± 16.35
1.87 ± 1.59
-
Ours w/o Smooth Loss
28.01 ± 29.23
17.73 ± 30.79
18.74 ± 20.96
5.72 ± 7.39
14.37 ± 15.89
5.64 ± 9.17
10.60 ± 4.61
Ours w/o Init Loss
9.35 ± 13.50
19.58 ± 39.17
14.64 ± 19.90
0.75 ± 0.90
1.90 ± 3.11
1.14 ± 2.60
13.07 ± 7.26
Ours w/o Noise Resistance
4.75 ± 7.83
2.22 ± 6.89
1.30 ± 1.09
0.17 ± 0.15
0.48 ± 0.79
1.10 ± 2.72
22.65 ± 3.13
Ours w/o Blended Rendering
1.72 ± 2.38
1.88 ± 6.23
1.88 ± 2.54
0.12 ± 0.07
0.34 ± 0.52
1.05 ± 2.61
22.23 ± 2.64
Table 1: Results and Ablation Studies on FreeArt-21. Metrics are reported as mean ± std over all test videos. Lower ( ↓ ) is better for the joint and geometry metrics, higher ( ↑ ) is better for the rendering metric. The best results are highlighted in bold .
Figure 3: Qualitative Results on FreeArt-21. We visualize both the articulation and rendering results of our method. The red part and the blue part are the identified parts by our method.
Joint Type
Method
Axis (deg) ↓
Position (cm) ↓
State (deg/cm) ↓
Revolute
Articulate-Anything [ 16 ]
46.98 ± 45.27
81.00 ± 40.00
N/A
RSRD [ 15 ]
67.06 ± 29.22
203.00 ± 748.00
59.02 ± 34.38
Video2Articulation [ 24 ]
18.34 ± 32.09
13.00 ± 25.00
14.32 ± 26.35
Ours
1.77 ± 2.87
1.31 ± 1.81
3.69 ± 6.60
Prismatic
Articulate-Anything [ 16 ]
52.71 ± 44.69
-
N/A
RSRD [ 15 ]
69.91 ± 24.07
-
70.00 ± 48.00
Table 2: Results on Video2Articulation-S Dataset. We report joint estimation results (top) and geometry reconstruction results (bottom). The best results are in bold .
Figure 4: Qualitative Results on Real-world Objects. Our method successfully reconstructs all the objects with correct joints, geometries, and textures.
Objects
Axis (deg) ↓
Position (cm) ↓
CD (cm) ↓
PSNR (dB) ↑
Drawer
1.06
-
1.91
19.82
Bin
4.68
0.20
3.46
23.75
Case
0.10
0.10
1.89
22.77
Lid
8.92
0.40
3.36
22.39
Fan
1.53
1.30
1.78
21.95
Cabinet
0.10
3.30
2.53
23.71
Table 3: Quantitative Results on Real-world Objects. The rotation and translation errors are clipped to 0.1∘ and 0.1 cm, respectively, which correspond to the smallest annotation units.
Figure 5: Illustration of Part Solver Module.
Joint Type
Method
Axis(deg) ↓
Position(cm) ↓
State(deg/cm) ↓
CD-w(cm) ↓
CD-m(cm) ↓
CD-s(cm) ↓
PSNR(dB) ↑
Revolute
2% Depth Noise
1.41 ± 1.49
0.51 ± 0.58
1.22 ± 1.27
0.23 ± 0.30
0.30 ± 0.27
1.05 ± 2.47
23.52 ± 4.65
AllTracker → CoTracker3
0.88 ± 0.84
0.46 ± 0.86
1.68 ± 1.30
0.79 ± 2.19
0.80 ± 1.08
1.47 ± 2.43
23.87 ± 4.94
DINOv3 → DINOv2
1.64 ± 1.22
0.32 ± 0.47
1.21 ± 0.77
0.11 ± 0.11
0.28 ± 0.27
1.04 ± 2.52
23.74 ± 4.41
Ours
1.04 ± 1.03
0.29 ± 0.36
1.43 ± 1.20
0.14 ± 0.13
0.28 ± 0.36
0.97 ± 2.69
24.02 ± 3.15
Prismatic
2% Depth Noise
1.86 ± 2.51
-
0.92 ± 0.66
0.34 ± 0.13
1.07 ± 1.18
1.78 ± 3.03
21.75 ± 2.99
AllTracker → CoTracker3
1.98 ± 3.09
-
1.12 ± 0.68
0.79 ± 1.08
0.98 ± 1.22
0.94 ± 1.18
21.85 ± 3.64
Table 4: Results of Robustness Analysis.
Method
Rot. Error (deg) ↓
Pos. Error (cm) ↓
Video2Articulation* [ 24 ]
4.621
9.246
Ours
2.483
1.526
Table 5: Results of Camera Pose Estimation on the Video2Articulation-S Dataset. * indicates that the results are taken from the original paper.
Figure 6: Visualization of camera pose trajectories in FreeArt-21. The black line shows the ground truth camera trajectory, and the yellow line shows our estimated camera trajectory.
Method / mIoU (%) ↑
FreeArt-21
Video2Articulation-S
Video2Articulation [ 24 ]
34.61
59.65*
Ours
90.88
84.92
Table 6: Results of Part Segmentation. * indicates that the results are taken from the original paper.
Figure 7: Visualization of all objects in FreeArt-21.
Figure 8: Visualization of our reconstruction results on FreeArt-21.
Articulated object reconstruction from sparse-view images is an ill-posed problem that requires simultaneous inference of geometry and underlying articulation structure. Existing methods for articulated object reconstruction based on NeRF and 3D Gaussian Splatting (3DGS) typically rely on dense views or strong priors (e.g., depth maps, joint types, predefined number of joints) and require costly per-object optimization. In this paper, we propose ArtSplat, the first feed-forward framework for articulated 3D Gaussian Splatting. It reconstructs both geometry and joint parameters from sparse multi-view images across multiple articulation states in a single forward pass. To address the challenges of single-pass articulated reconstruction, we introduce a per-pixel joint map representation that enables the integration of joint parameter estimation into the feed-forward pipeline. We further propose a Cross-State Attention (CSA) mechanism with state tokens, which effectively captures discrete motion across input states. Experiments on 68 articulated objects from PartNet-Mobility, including both single- and multi-joint configurations, demonstrate that ArtSplat achieves competitive performance in both geometry and joint estimation, while being over 400 times faster than baselines.
Inseo Lee, Yoonji Kim, Eugene Sohn +4
1Seoul National University · 2Sogang University · 3NAVER AI Lab
Reconstructing articulated objects is essential for building digital twins of interactive environments. However, prior methods typically decouple geometry and motion by first reconstructing object shape in distinct states and then estimating articulation through post-hoc alignment. This separation complicates the reconstruction pipeline and restricts scalability, especially for objects with complex, multi-part articulation. We introduce a unified representation that jointly models geometry and motion using articulated 3D Gaussians. This formulation improves robustness in motion decomposition and supports articulated objects with up to 20 parts, significantly outperforming prior approaches that often struggle beyond 2--3 parts due to brittle initialization. To systematically assess scalability and generalization, we propose MPArt-90, a new benchmark consisting of 90 articulated objects across 20 categories, each with diverse part counts and motion configurations. Extensive experiments show that our method consistently achieves superior accuracy in part-level geometry reconstruction and motion estimation across a broad range of object types. We further demonstrate applicability to downstream tasks such as robotic simulation and human-scene interaction modeling, highlighting the potential of unified articulated representations in scalable physical modeling.
Reconstructing articulated objects with multiple movable parts is essential for understanding object structure and enabling physical interaction. However, this reconstruction task poses significant challenges due to the entanglement of geometry, appearance, and motion parameters during optimization. Existing methods rely primarily on photometric supervision, which commonly fails to disentangle these interdependent components, resulting in poor part decomposition with blurred boundaries and geometric artifacts. To address this limitation, we introduce StructureGS, a reconstruction framework for articulated objects that integrates structure-aware guidance into 3D Gaussian Splatting. Our approach leverages oriented bounding boxes of object parts to enforce two key structural properties: spatial coherence, which constrains each part's geometry to remain compact and spatially coherent within its designated region, and structural connectivity, which enforces physically plausible contact relationships between adjacent parts. These properties are realized through structure-aware losses that inject explicit structural constraints into the optimization process. Extensive experiments demonstrate that our method achieves state-of-the-art performance in articulated object reconstruction, producing high-quality results with well-defined part geometries.