AI Paper Atlas

Research map / 3D vision / Face and shape

MERG3R: A Divide-and-Conquer Approach to Large-Scale Neural Visual Geometry

Leo Kaixuan Cheng, Abdus Shaikh, Ruofan Liang, Zhijie Wu, Yushi Guan, Nandita Vijaykumar

CVPR 2026 · 3D vision › Face and shape · approach: Neural implicit surface reconstruction · 1 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
interleaved sampling when splitting the pseudo-video into subsets (vs graph clustering) (ablation)MERG3R + Pi3 · Tanks & TemplesRRE: 1.280 -> 0.178Improved
interleaved sampling vs sliding window split (ablation)MERG3R + Pi3 · Tanks & TemplesRRE: 0.492 -> 0.178Improved
add global bundle adjustment on merged tracks (ablation)MERG3R + VGGT (subset 50) · Tanks & TemplesRRE: 2.284 -> 1.563Improved
use LightGlue graph tracking instead of VGGT tracking for BA (ablation)MERG3R + VGGT · Tanks & TemplesRRE / memory: RRE 2.659 -> 1.563; memory 33.22 -> 12.68 GBImproved
VGGT tracking for BA instead of no BA (ablation)MERG3R + VGGT · Tanks & TemplesATE: 0.551 -> 0.602Hurt
graph-based matching rather than adjacent-frame matching of the pseudo video (ablation)MERG3R + VGGT · Tanks & TemplesRRE: 1.572 -> 1.563No clear effect
larger subset size (ablation)MERG3R + Pi3 · Tanks & TemplesRRE: 0.674 (25) -> 0.177 (100)Improved
larger subset size beyond 100 (ablation)MERG3R + Pi3 · Tanks & Templestime and memory: 136.51 s / 11.63 GB (100) -> 155.39 s / 14.31 GB (200)Hurt
larger overlap between subsets (ablation)MERG3R + Pi3 · Tanks & TemplesRRE: 0.198 (3) -> 0.162 (15)No clear effect
use pseudo-video ordering instead of real video order (ablation)MERG3R · Tanks & TemplesATE: Pi3 RRE 0.196 -> 0.178No clear effect
wrap base models with MERG3R divide-and-conquerVGGT*, FastVGGT, Pi3 · 7-Scenes (1000 images)AUC@30: OOM -> 82.20 / 81.45 / 83.63Improved
wrap Pi3 with MERG3RPi3 · 7-Scenes (500 images)AUC@30: 83.89 -> 82.97No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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