AI Paper Atlas

Research map / 3D vision / Face and shape

SE(n)-Invariant Flow Matching: A General Framework with Application to Object Reassembly

Gaël Heck, Sylvie Le Hégarat-Mascle, Nicolas Lermé

ICML 2026 · 3D vision › Face and shape · approach: Articulated and non-rigid shape learning

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
add canonical gauge fixing (ablation)SE(2) flow matching · SVJPOverlap: .80 -> .90Improved
add quotient-invariant loss (ablation)SE(2) flow matching · SVJPOverlap: .80 -> .92Improved
add quotient loss on top of gauge fixing (ablation)SE(2) flow matching · SVJPOverlap: .90 -> .93Improved
replace gauge fixing with multi-anchor supervision (ablation)SE(2) flow matching · SVJPOverlap: .90 -> .87Hurt
add multi-anchor supervision to base (ablation)SE(2) flow matching · SVJPOverlap: .80 -> .87Improved
add test-time prediction projection (ablation)SE(2) flow matching · SVJPOverlap: .93 -> .94Improved
add gauge fixing (ablation)SE(3) flow matching · Breaking Bad EverydayRMSE(R): 9.67 -> 6.98Improved
add quotient loss (ablation)SE(3) flow matching · Breaking Bad EverydayRMSE(R): 9.67 -> 6.07Improved
add gauge fixing and quotient loss (ablation)SE(3) flow matching · Breaking Bad ArtifactRMSE(R): 10.02 -> 5.41Improved
use invariant flow matching instead of strongest SE(2) baseline at 1% noiseSVFP · SVFPOverlap: 0.76 to 0.83Improved
use invariant flow matching instead of GARFSE(3) reassembly · Breaking Bad EverydayRMSE(R): 6.43 to 5.80Improved
use invariant flow matching instead of Islam et al.RePAIR · RePAIROverlap: 0.21 -> 0.41Improved

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