AI Paper Atlas

Research map / 3D vision / 3D generation

SceneVerse: Scaling 3D Vision-Language Learning for Grounded Scene Understanding

Baoxiong Jia, Yixin Chen, Huangyue Yu, Yan Wang, Xuesong Niu, Tengyu Liu, Qing Li, Siyuan Huang

ECCV 2024 · 3D vision › 3D generation · approach: Open-vocabulary 3D scene understanding · 90 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
pre-train on SceneVerse before fine-tuningGPS · ScanRefer Acc@0.5 overallaccuracy: 40.4 -> 47.1 (pre-train) -> 48.1 (fine-tuned)Improved
train GPS from scratch on benchmark data only (ablation)GPS · Nr3Doverall accuracy: 58.7 vs 3D-VisTA 64.2Hurt
zero-shot transfer with SceneVerse pre-training versus training from scratchGPS · SceneVerse-valoverall accuracy: 38.5 -> 59.2Improved
zero-shot GPS versus 3D-VisTA on existing benchmarksGPS · Nr3D / Sr3D / ScanReferaccuracy: Nr3D 35.2 -> 32.4; Sr3D 31.2 -> 33.3Mixed
add generated template text to pre-training (ablation)GPS · ScanRefer Acc@0.5accuracy: 38.4 -> 46.1Improved
add LLM-rephrased text on top of templates (ablation)GPS · ScanRefer Acc@0.5accuracy: 46.1 -> 46.3No clear effect
add human-annotated text on top of generated text (ablation)GPS · ScanRefer Acc@0.5accuracy: 46.3 -> 47.1Improved
add scene-level alignment loss (ablation)GPS · SceneVerse-valoverall accuracy: 62.4 -> 66.9Improved
add masked language modeling loss without scene-level loss (ablation)GPS · SceneVerse-valoverall accuracy: 65.2 -> 62.4Hurt
pre-train on synthetic scenes only (Structured3D or ProcTHOR) then test on real scenes (ablation)GPS · SceneVerse-valoverall accuracy: 64.8 (real) -> 7.0 (S3D) and 4.2 (ProcTHOR)Hurt
pre-train RegionPLC on SceneVerse dataRegionPLC · ScanNet open-vocab segmentation, SPUNet32mIoU: 59.6 -> 61.0Improved
scale pre-training data from 1/8 to full SceneVerse (ablation)GPS · ScanRefer and SceneVerse-valaccuracyImproved

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