Research map / Point clouds / Shape and surface completion
Point Scene Understanding via Disentangled Instance Mesh Reconstruction
Jiaxiang Tang, Xiaokang Chen, Jingbo Wang, Gang Zeng
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 residual bounding box learning (ResBox) (ablation) | DIMR · ScanNet mesh completion | IoU@0.25 mAP: 41.76 -> 46.34 | Improved |
| add multi-scale clustering (MSC) at test time (ablation) | DIMR · ScanNet mesh completion | IoU@0.25 mAP: 43.57 -> 46.34 | Improved |
| add ICP post-processing of meshes (ablation) | DIMR · ScanNet mesh completion | IoU@0.25 mAP: 42.33 -> 46.34 | Improved |
| add ICP post-processing of meshes (ablation) | DIMR · ScanNet mesh completion | CD@0.1 mAP: 52.19 -> 52.39 | No clear effect |
| add multi-scale clustering (MSC) (ablation) | DIMR · ScanNet mesh completion | CD@0.1 mAP: 48.26 -> 52.39 | Improved |
| replace detection backbone with segmentation backbone (Seg + Entangled vs Det + Entangled) (ablation) | RfD-Net · ScanNet mesh completion | IoU@0.25 mAP: 42.52 -> 43.95 | Improved |
| replace entangled occupancy learning with disentangled mesh reconstruction (DIMR) (ablation) | Seg backbone · ScanNet mesh completion | IoU@0.25 mAP: 43.95 -> 46.34 | Improved |
| replace entangled occupancy learning with disentangled mesh reconstruction (DIMR) (ablation) | Seg backbone · ScanNet mesh completion | CD@0.1 mAP: 47.46 -> 52.39 | Improved |
| use segmentation backbone instead of detection to reduce false positives | RfD-Net · ScanNet object recognition | Precision@0.25: 22.99 -> 43.70 | Improved |
| use DIMR instead of RfD-Net at high IoU threshold | mesh completion · ScanNet mesh completion | IoU@0.5 mAP: 14.35 -> 12.54 | Hurt |
| add projection or retrieval from an external model pool | DIMR · ScanNet point-to-mesh mapping | mean AP: 56.76 -> 58.03 | Improved |
| use generated meshes instead of retrieved CAD models on sofa | DIMR · ScanNet point-to-mesh mapping | sofa AP: 61.00 -> 60.33 | Mixed |
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.
- RfD-Net — RfD-Net: Point Scene Understanding by Semantic Instance Reconstruction (CVPR 2021) · better on 9 of 10 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
- RfD-Net: Point Scene Understanding by Semantic Instance Reconstruction (CVPR 2021 · similarity 0.92)
- DOCTR: Disentangled Object-Centric Transformer for Point Scene Understanding (AAAI 2024 · similarity 0.90)
- Point-based Instance Completion with Scene Constraints (ICLR 2025 · similarity 0.89)
- Know Your Neighbors: Improving Single-View Reconstruction via Spatial Vision-Language Reasoning (CVPR 2024 · similarity 0.88)
- MESC-3D:Mining Effective Semantic Cues for 3D Reconstruction from a Single Image (CVPR 2025 · similarity 0.88)
- SPAR3D: Stable Point-Aware Reconstruction of 3D Objects from Single Images (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.
