AI Paper Atlas

Research map / 3D vision / 3D generation

Lifting Unlabeled Internet-level Data for 3D Scene Understanding

Yixin Chen, Yaowei Zhang, Huangyue Yu, Junchao He, Yan Wang, Jiangyong Huang, Hongyu Shen, Junfeng Ni, Shaofei Wang, Baoxiong Jia, Song-Chun Zhu, Siyuan Huang

CVPR 2026 · 3D vision › 3D generation · approach: Open-vocabulary 3D scene understanding · 0 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
Pretrain on SceneVerse++ then finetune on ScanNetSpatialLM · ScanNet detectionF1@0.25: 38.0 -> 58.6Improved
Pretrain on SceneVerse++ instead of synthetic data (zero-shot)SpatialLM · ScanNet detectionF1@0.25: 29.0 -> 30.9Improved
Pretrain on SceneVerse++ instead of synthetic data (zero-shot)SpatialLM · ARKitScenes detectionF1@0.5: 21.2 -> 20.7No clear effect
Pretrain on SceneVerse++ only, zero-shotMask3D · ScanNet instance segmentationAP25: 36.1 -> 15.4Hurt
Pretrain on SceneVerse++ then finetune on ScanNetMask3D · ScanNet instance segmentationAP25: 36.1 -> 38.5Improved
Train on SceneVerse++ spatial QAQwen2.5-VL-3B · VSI-Bench full setAverage score: 27.9 -> 42.8Improved
Train on SceneVerse++ spatial QAQwen2.5-VL-7B · VSI-Bench full setAverage score: 36.6 -> 46.4Improved
Combine SceneVerse++ with SN/SN++ data (vs SN/SN++ only)Qwen2.5-VL-3B · VSI-Bench ARKit subsetAverage score: 49.0 -> 51.3Improved
Train on SceneVerse++ instead of SN/SN++Qwen2.5-VL-7B · VSI-Bench ARKit subsetAverage score: 48.8 -> 49.1No clear effect
Pretrain on SceneVerse++ then finetune on R2R (vs R2R only)LLaVA-Video · R2RSuccess rate: 0.088 -> 0.228Improved
Pretrain then finetune instead of mixing SceneVerse++ with R2RLLaVA-Video · R2RSuccess rate: 0.188 -> 0.228Improved
Add trajectory refinement (ablation)SceneVerse++ VLN data · R2RSuccess rate: 0.177 -> 0.228Improved

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