cat

Can 3D Generative Models Help 3D Assembly?

New York University
*, † Equal Contribution, ✉ Corresponding Author

CRAG shows that assembly and generation are mutually reinforcing: assembly provides part-level structural priors for generation, while generation injects holistic shape context that resolves ambiguities in assembly.

Bowl
Plate
Pot
Vase
Vase
Bowl
Plate
Pot
Vase
Vase

Note: All fragments shown above are obtained from real 3D scans. Even when presented with unfamiliar fragments or missing parts, GARF successfully assembles them and accurately aligns their poses. Try rotating and zooming in on the 3D models to see how seamlessly the pieces fit together!All the fragments above are obtained from real scans.

Abstract

Most existing 3D assembly methods treat the problem as pure pose estimation, rearranging observed parts via rigid transformations. In contrast, human assembly naturally couples structural reasoning with holistic shape inference. Inspired by this intuition, we reformulate 3D assembly as a joint problem of assembly and generation. We show that these two processes are mutually reinforcing: assembly provides part-level structural priors for generation, while generation injects holistic shape context that resolves ambiguities in assembly. Unlike prior methods that cannot synthesize missing geometry, we propose CRAG, which simultaneously generates plausible complete shapes and predicts poses for input parts. Extensive experiments demonstrate state-of-the-art performance across in-the-wild objects with diverse geometries, varying part counts, and missing pieces.

Methodology

Overview Image

Overall illustration of CRAG. Our model consists of two interacting branches: an Assembly Branch that predicts the pose for each part via SE(3) flow matching, and a Generation Branch that synthesizes the complete shape via flow matching. A Joint Adapter bridges these branches, enabling bidirectional information flow. We employ a two-stage training strategy: learning assembly first, and then jointly finetuning both tasks

Assembly Results

Explore our 3D reassembly results interactively. The models below demonstrate how GARF accurately aligns fragments across different material categories. You can rotate, zoom, and inspect the reassembled objects from any angle.

Breaking Bad Dataset

Everyday
Everyday
Everyday
Everyday
Artifact
Artifact
Artifact
Artifact

Fractura Dataset (Synthetic Fracture Subset)

Following the simulated fracture method used in Breaking Bad, we also performed simulated breakage on the complete objects in our Fractura dataset.

Fantastic Break Dataset

Although the objects in the Fantastic Break dataset only break into two pieces, the challenge lies in the fact that the fracture surfaces are obtained from real scans, not simulations.

BibTeX

@inproceedings{li2025garf,
 title={GARF: Learning Generalizable 3D Reassembly for Real-World Fractures},
 author={Li, Sihang and Jiang, Zeyu and Chen, Grace and Xu, Chenyang and Tan, Siqi and Wang, Xue and Fang, Irving and Zyskowski, Kristof and McPherron, Shannon P and Iovita, Radu and Feng, Chen and Zhang, Jing},
 year={2025},
 booktitle={International Conference on Computer Vision (ICCV)}
}

Acknowledgements

We gratefully acknowledge the Physical Anthropology Unit, Universidad Complutense de Madrid for access to curated human skeletons, and Dr. Scott A. Williams (NYU Anthropology Department) for the processed data samples. This work was supported in part through NSF grants 2152565, 2238968, 2322242, and 2426993, and the NYU IT High Performance Computing resources, services, and staff expertise.