AI Paper Atlas

Research map / 3D vision / 3D generation

DepWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors

Seok-Young Kim, Abdelrahman Elskhawy, TAEWOOK HA, Dooyoung Kim, Eunjae Shin, Benjamin Busam, Woontack Woo

ECCV 2026 · 3D vision › 3D generation · approach: Open-vocabulary 3D scene understanding

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 box-conditioned instance masking for depth sampling (ablation)2D lifting baseline · 3DSSGRelationship recall: 28.5 -> 40.3Improved
add box-conditioned instance masking for depth sampling (ablation)2D lifting baseline · 3DSSGObject recall: 63.4 -> 71.7Improved
add dual-domain depth refinement (DR) (ablation)DeWorldSG with instance mask · 3DSSGRelationship recall: 40.3 -> 47.8Improved
add dual-domain depth refinement (DR) (ablation)DeWorldSG with instance mask · 3DSSGPredicate recall: 46.7 -> 52.9Improved
add V-JEPA 2 world-model prior (ablation)DeWorldSG with mask and DR · 3DSSGRelationship recall: 47.8 -> 50.2Improved
add V-JEPA 2 world-model prior (ablation)DeWorldSG with mask and DR · 3DSSGPredicate mRecall: 30.1 -> 36.0Improved
add V-JEPA 2 world-model prior (ablation)DeWorldSG with mask and DR · 3DSSGObject mRecall: 70.8 -> 70.6No clear effect
use V-JEPA 2 temporal prior instead of DINOv2 static prior (ablation)DeWorldSG · 3DSSGRelationship recall: 49.5 -> 50.2Improved
use V-JEPA 2 temporal prior instead of DINOv2 static prior (ablation)DeWorldSG · 3DSSGPredicate mRecall: 33.7 -> 36.0Improved
add DINOv2 prior instead of no prior (ablation)DeWorldSG with mask and DR · 3DSSGPredicate recall: 52.9 -> 56.6Improved
replace GT camera poses with ORB-SLAM3 estimated posesDeWorldSG · ReplicaSSGRelationship recall: 38.2 -> 37.4Hurt
replace GT camera poses with ORB-SLAM3 estimated posesDeWorldSG · ReplicaSSGObject recall: 35.4 -> 33.8Hurt

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