AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

EG-3DVG: Expression and Geometry Aware Grounding Decoder for 3D Visual Grounding

GwangWook Park, Hyo-Jun Lee, Jong-Hyeon Baek, Hanul Kim, Yeong Jun Koh

CVPR 2026 · Vision-language and multimodal › Human and embodied modeling · approach: Visual grounding transformers

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 PECA, GMA and ECL together (ablation)EG-3DVG baseline decoder · ScanRefer3DREC Acc@0.5: 49.62 -> 52.36Improved
add PECA alone (ablation)baseline decoder · ScanRefer3DRES mIoU: 44.19 -> 45.69Improved
add GMA alone (ablation)baseline decoder · ScanRefer3DRES mIoU: 44.19 -> 46.19Improved
add ECL alone (ablation)baseline decoder · ScanRefer3DRES mIoU: 44.19 -> 45.84Improved
add GMA alone (ablation)baseline decoder · ScanRefer3DREC Acc@0.25: 57.04 -> 57.05No clear effect
add ECL on top of PECA and GMA (ablation)EG-3DVG · ScanRefer3DRES mIoU: 46.57 -> 47.28Improved
add GMA on top of PECA and ECL (ablation)EG-3DVG · ScanRefer3DREC Acc@0.5: 51.64 -> 52.36Improved
use text-and-geometry attention mask instead of fixed-radius mask in GMA (ablation)GMA · ScanRefer3DREC Acc@0.5: 51.12 -> 52.36Improved
include text tokens in the GMA attention mask instead of geometric feature only (ablation)GMA · ScanRefer3DRES mIoU: 46.70 -> 47.28Improved
EG-3DVG vs previous best single-stage TSP3DTSP3D · ScanRefer single-stageAcc@0.5: +3.36Improved
EG-3DVG vs MCLN in two-stage settingMCLN · ScanRefer two-stageAcc@0.5: +6.83Improved
EG-3DVG vs G3-LQ on SR3D hard subsetG3-LQ · SR3D hardAcc@0.25: 66.3 -> 66.6No clear effect

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