AI Paper Atlas

Research map / 3D vision / Face and shape

GARF: Learning Generalizable 3D Reassembly for Real-World Fractures

Sihang Li, Zeyu Jiang, Grace Chen, Chenyang Xu, Siqi Tan, Xue Wang, Irving Fang, Kristof Zyskowski, Shannon P. McPherron, Radu Iovita, Chen Feng, Jing Zhang

ICCV 2025 · 3D vision › Face and shape · approach: Articulated and non-rigid shape learning · 10 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace VAE-based pretraining with fracture-aware pretraining (ablation)PuzzleFusion++ · Breaking Bad EverydayRMSE(R) (deg): 32.91 -> 15.36Improved
add large-fragment attention and weighted sampling with flow matching (ablation)PuzzleFusion++ + pretraining · Breaking Bad EverydayRMSE(R) (deg): 15.55 -> 8.70Improved
use flow matching instead of diffusion as denoiser (ablation)PuzzleFusion++ with pretraining and large-fragment attention · Breaking Bad EverydayRMSE(R) (deg): 12.58 -> 8.70Improved
use flow matching instead of diffusion without large-fragment attention (ablation)PuzzleFusion++ with pretraining · Breaking Bad EverydayRMSE(R) (deg): 15.36 -> 15.55No clear effect
add two-session flow matching (one-step init + multi-step) (ablation)GARF-mini · Breaking Bad EverydayRMSE(R) (deg): 7.63 -> 6.68Improved
replace PointNet++ with PTv3 backbone (ablation)GARF-mini · Breaking Bad EverydayRMSE(R) (deg): 7.40 -> 6.68Improved
add multi-anchor training (ablation)GARF-mini flow matching · Breaking Bad EverydayRMSE(R) (deg): 8.02 -> 7.63Improved
use anchor fragments instead of anchor-free training (ablation)GARF · Breaking Bad EverydayRMSE(R) (deg): 9.09 -> 6.10Improved
use random anchor instead of largest anchor (ablation)GARF · Breaking Bad EverydayRMSE(R) (deg): 6.10 -> 6.09No clear effect
increase single-session flow steps from 20 to 50 (ablation)GARF-mini · Breaking Bad EverydayRMSE(R) (deg): 7.63 -> 7.50No clear effect
add 20% extraneous fragmentsGARF · Breaking Bad Everyday, 5-17 fragmentsPA (%): 83.39 -> 79.21Hurt

Later papers that compared with this method

3 later accepted papers report GARF as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy