AI Paper Atlas

Research map / 3D vision / Depth and stereo

RoMa v2: Harder Better Faster Denser Feature Matching

Johan Edstedt, David Nordström, Yushan Zhang, Georg Bökman, Jonathan Astermark, Viktor Larsson, Anders Heyden, Fredrik Kahl, Mårten Wadenbäck, Michael Felsberg

ECCV 2026 · 3D vision › Depth and stereo · approach: Dense feature matching for correspondence · 55 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
add refiner EMA (decay 0.999) (ablation)RoMa v2 refiners · MegaDepth-1500AUC@5deg: 61.4 -> 62.8Improved
add refiner EMA (ablation)RoMa v2 refiners · ScanNet-1500AUC@5deg: 33.6 -> 33.6No clear effect
use DINOv3 instead of DINOv2 in coarse matcher (ablation)RoMa v2 coarse matcher · WxBSscore: 35.6 -> 34.2Mixed
use DINOv3 instead of DINOv2 features (linear probe) (ablation)frozen encoder · MegaDepth linear probeEPE: 27.1 -> 19.0Improved
replace UFM matching objective with RoMa v2 attention plus NLL objective (ablation)coarse matcher · Hypersim PCK@1pxPCK: 11.2 -> 30.5Improved
replace UFM training/architecture setup with RoMa v2 (ablation)matcher · Hypersim PCK@5pxPCK: 67.4 -> 86.7Improved
use custom CUDA local correlation kernel (ablation)RoMa v2 refiners · runtimememory (GB): 5.6 -> 4.8Improved
use CUDA kernel for local correlation (ablation)RoMa v2 refiners · runtimethroughput (pairs/s): 30.3 -> 30.9No clear effect
simulate less precise subpixel accuracy by perturbing predictions (ablation)RoMa v2 · MegaDepth-1500AUC@5deg: 62.8 -> 46.3Hurt
upgrade RoMa to RoMa v2 (new matcher, data, refiners)RoMa · Map-freePrecision: 59.7 -> 73.3Improved
upgrade RoMa to RoMa v2RoMa · AerialMegaDepthEPE: 25.05 -> 4.12Improved
upgrade RoMa to RoMa v2RoMa · WxBSmAA@10px: 60.8 -> 55.4Hurt

Later papers that compared with this method

1 later accepted paper reports RoMa v2 as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.

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