Research map / 3D vision / 3D generation
Towards Part-Based Understanding of RGB-D Scans
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| guide part-mask prediction with a predicted part tree and geometric priors | MLCVNet object detections · Scan2CAD semantic part completion | IoU (class avg): 11.9 -> 16.7 | Helped |
| guide part-mask prediction with a predicted part tree and geometric priors | MLCVNet object detections · Scan2CAD semantic part completion | Chamfer Distance (class avg): 0.068 -> 0.062 | Helped |
| use part-tree message passing (ablation) | Ours · Scan2CAD semantic part completion | IoU (class avg): 14.3 -> 16.7 | Helped |
| add geometric part priors (ablation) | Ours · Scan2CAD semantic part completion | IoU (class avg): 15.4 -> 16.7 | Helped |
| add prior refinement (ablation) | Ours · Scan2CAD semantic part completion | IoU (class avg): 16.4 -> 16.7 | Helped |
| refine priors with residual offsets rather than absolute predictions (ablation) | Ours · Scan2CAD semantic part completion | IoU (class avg): 15.8 -> 16.7 | Helped |
| use voxel resolution 16 rather than 32 (ablation) | Ours · Scan2CAD semantic part completion | IoU (class avg): 16.7 -> 18.8 | Helped |
| use voxel resolution 64 rather than 32 (ablation) | Ours · Scan2CAD semantic part completion | IoU (class avg): 16.7 -> 13.1 | Hurt |
| use voxel resolution 16 rather than 32 (ablation) | Ours · Scan2CAD semantic part completion | Chamfer Distance (class avg): 0.062 -> 0.070 | Hurt |
| use voxel resolution 64 rather than 32 (ablation) | Ours · Scan2CAD semantic part completion | Chamfer Distance (class avg): 0.062 -> 0.069 | Hurt |
| predict complete object geometry through part decomposition | MLCVNet object detections · Scan2CAD instance completion | IoU (avg): 35.9 -> 39.3 | Helped |
| predict complete object geometry through part decomposition | MLCVNet object detections · Scan2CAD instance completion | mAP@25 (avg): 55.3 -> 64.8 | Helped |
Closest papers by meaning
- PartNeXt: A Next-Generation Dataset for Fine-Grained and Hierarchical 3D Part Understanding (NeurIPS 2025 · similarity 0.92)
- PartNet: A Large-Scale Benchmark for Fine-Grained and Hierarchical Part-Level 3D Object Understanding (CVPR 2019 · similarity 0.91)
- A Unified Framework for 3D Scene Understanding (NeurIPS 2024 · similarity 0.89)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · similarity 0.89)
- PARTONOMY: Large Multimodal Models with Part-Level Visual Understanding (NeurIPS 2025 · similarity 0.88)
- Edge-Aware 3D Instance Segmentation Network with Intelligent Semantic Prior (CVPR 2024 · similarity 0.88)
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.