Research map / 3D vision / 3D generation
Articulated Object Reconstruction from Rest-State Observation
Daeun Lee, Jaeah Lee, Woosung Kim, Haebeom Jung, Jaesik Park
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add co-refinement loop between GPT-5.2 and SAM3 (ablation) | GPT-5.2 + SAM3 · ACD | hierarchy tree accuracy: 62.3 -> 72.5 | Improved |
| add co-refinement loop between Qwen3-VL and SAM3 (ablation) | Qwen3-VL + SAM3 · ACD | hierarchy tree accuracy: 52.2 -> 72.5 | Improved |
| add co-refinement loop between Qwen3-VL and SAM3 (ablation) | Qwen3-VL + SAM3 · ACD | per-part count accuracy: 47.8 -> 63.8 | Improved |
| add co-refinement loop between GPT-5.2 and SAM3 (ablation) | GPT-5.2 + SAM3 · ACD | per-part count accuracy: 59.4 -> 59.4 | No clear effect |
| add co-refinement to Singapo (ablation) | Singapo · ACD-ABO | type accuracy: 41.47 -> 67.94 | Improved |
| add co-refinement to Singapo (ablation) | Singapo · ACD-HSSD | type accuracy: 59.42 -> 61.21 | Improved |
| use 2DGS instead of 3DGS for mesh backend (ablation) | proposed pipeline · ACD | segmentation IoU: 0.30 -> 0.97 | Improved |
| use 2DGS instead of 3DGS for mesh backend (ablation) | proposed pipeline · ACD | axis direction error: 1.69 -> 3.67 | Hurt |
| add mask overlay to video generation (ablation) | Wan2.2 + VBVR LoRA · ACD-HSSD | accurate articulation rate: 69.2% -> 80.8% | Improved |
| add mask overlay to video generation (ablation) | Wan2.2 + VBVR LoRA · ACD-HSSD | hallucination rate: 23.1% -> 11.5% | Improved |
| raise max refinement rounds from 2 to 20 (ablation) | co-refinement · ACD | convergence rate: 84.1% -> 92.8% | Improved |
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.
- REArtGS++ — REArtGS: Reconstructing and Generating Articulated Objects via 3D Gaussian Splatting with Geometric and Motion Constraints (NeurIPS 2025) · better on 2 of 3 shared results
- PartField — PartField: Learning 3D Feature Fields for Part Segmentation and Beyond (ICCV 2025) · better on 1 of 1 shared result
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
- DIPO: Dual-State Images Controlled Articulated Object Generation Powered by Diverse Data (NeurIPS 2025 · similarity 0.90)
- Real2Code: Reconstruct Articulated Objects via Code Generation (ICLR 2025 · similarity 0.90)
- Building Interactable Replicas of Complex Articulated Objects via Gaussian Splatting (ICLR 2025 · similarity 0.90)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.90)
- SINGAPO: Single Image Controlled Generation of Articulated Parts in Objects (ICLR 2025 · similarity 0.89)
- ArticulatedGS: Self-supervised Digital Twin Modeling of Articulated Objects using 3D Gaussian Splatting (CVPR 2025 · similarity 0.89)
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.
