Research map / 3D vision / Face and shape
JRM: Joint Reconstruction Model for Multiple Objects without Alignment
Qirui Wu, Yawar Siddiqui, Duncan Frost, Samir Aroudj, Armen Avetisyan, Richard Newcombe, Angel X. Chang, Jakob Engel, Henry Howard-Jenkins
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 joint coupled reconstruction (JRM) over independent FM with 3 rescans | JRM · Temporal repetition, 3 rescans | CD: 4.43 -> 3.07 (vs FM aligned by MORE2) | Improved |
| add more rescans for FM with estimated alignment | FM · Temporal repetition | CD: 3.07 -> 4.43 (target only to 3 rescans) | Hurt |
| add more rescans for JRM | JRM · Temporal repetition | CD: 3.46 -> 2.95 (target only to 1 rescan) | Improved |
| JRM vs FM with target scan only | JRM · Temporal repetition, target only | CD: 3.46 vs 3.07 | No clear effect |
| add an identical source object | JRM · Spatial repetition | CD: 2.61 -> 2.49 | Improved |
| add a negative (different) source object to FM | FM · Spatial repetition | CD: 3.33 -> 8.83 | Hurt |
| add a negative source object to JRM | JRM · Spatial repetition | CD: 2.61 -> 3.04 | Hurt |
| use JRM instead of FM with oracle alignment for articulated objects | JRM · Articulation state 1 | CD: 5.37 -> 3.66 | Improved |
| use JRM instead of FM on real scenes | JRM · ScanNet++ | CD: 4.20 -> 2.69 | Improved |
| use JRM instead of FM on real scenes | JRM · Replica | CD: 3.74 -> 3.21 | Improved |
| train only on similar pairs (neg ratio 0 vs 0.1) (ablation) | JRM · Spatial repetition, negative pair | CD: 6.22 vs 2.56 | Hurt |
| train only on negative pairs (ratio 1 vs 0.1) (ablation) | JRM · Spatial repetition, identical pair | CD: 2.18 -> 2.69 | Hurt |
Closest papers by meaning
- One Algorithm to Align Them All (CVPR 2026 · similarity 0.87)
- Reconstruction by Generation: 3D Multi-Object Scene Reconstruction from Sparse Observations (ECCV 2026 · similarity 0.87)
- CUPID: Generative 3D Reconstruction via Joint Object and Pose Modeling (CVPR 2026 · similarity 0.87)
- Articulated Object Reconstruction from Rest-State Observation (ECCV 2026 · similarity 0.87)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (ICLR 2024 · similarity 0.86)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.85)
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.
