Research map / 3D vision / Face and shape
MessyKitchens: Contact-rich object-level 3D scene reconstruction
Junaid Ahmed Ansari, Ran Ding, Fabio Pizzati, Ivan Laptev
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 Multi-Object Decoder to SAM 3D | SAM 3D · MessyKitchens (scene-level) | IoU: 0.431 -> 0.472 | Improved |
| add Multi-Object Decoder to SAM 3D | SAM 3D · HouseCat6D (object-level) | IoU: 0.325 -> 0.404 | Improved |
| add Multi-Object Decoder to SAM 3D | SAM 3D · GraspNet-1B (object-level) | IoU: 0.336 -> 0.344 | Improved |
| add Multi-Object Decoder to SAM 3D | SAM 3D · GraspClutter6D (scene-level) | IoU: 0.487 -> 0.496 | Improved |
| add Multi-Object Decoder to SAM 3D | SAM 3D · MessyKitchens contacts | contact F1 (%): 55.8 -> 66.1 | Improved |
| add multi-object attention on pose tokens to MOD with shape-only attention (ablation) | MOD · MessyKitchens (scene-level) | IoU: 0.463 -> 0.472 | Improved |
| add multi-object attention on shape tokens to MOD with pose-only attention (ablation) | MOD · MessyKitchens (scene-level) | IoU: 0.457 -> 0.472 | Improved |
| use K=3 blocks instead of K=1 (ablation) | MOD · MessyKitchens (scene-level) | IoU: 0.467 -> 0.472 | Improved |
| use K=6 blocks instead of K=3 (ablation) | MOD · MessyKitchens (scene-level) | IoU: 0.472 -> 0.427 | Hurt |
| use K=6 blocks instead of K=3 (ablation) | MOD · MessyKitchens (object-level) | IoU: 0.445 -> 0.403 | Hurt |
| add normal-consistency term to distance-only registration (ablation) | object-to-scene registration · MessyKitchens registration | depth error | Improved |
| train MOD on synthetic data only, evaluate zero-shot on other datasets | MOD · GraspClutter6D (object-level) | IoU: 0.328 -> 0.340 | Improved |
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.
- SAM 3D — SAM 3D: 3Dfy Anything in Images (CVPR 2026) · better on 17 of 17 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
- CoReNet: Coherent 3D Scene Reconstruction from a Single RGB Image (ECCV 2020 · similarity 0.88)
- DECO: Dense Estimation of 3D Human-Scene Contact In The Wild (ICCV 2023 · similarity 0.88)
- RHINO: Reconstructing Human Interactions with Novel Objects from Monocular Videos (CVPR 2026 · similarity 0.88)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.87)
- PICO: Reconstructing 3D People In Contact with Objects (CVPR 2025 · similarity 0.87)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.87)
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.
