Research map / 3D vision / Face and shape
From Points to Multi-Object 3D Reconstruction
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 |
|---|---|---|---|
| Use geometric-similarity soft labels instead of one-hot shape labels (ablation) | Points2Objects shape selection · ShapeNet-triplets | Abs. 3D IoU mean: 32.2 -> 36.4 | Helped |
| Use geometric-similarity soft labels instead of one-hot shape labels (ablation) | Points2Objects shape selection · ShapeNet-triplets | Abs. 3D IoU global: 40.3 -> 44.7 | Helped |
| Add the collision loss (ablation) | Points2Objects · ShapeNet-triplets validation | mean intersecting volume: 1168.8 -> 794 | Helped |
| Add the collision loss (ablation) | Points2Objects · ShapeNet-triplets validation | number of collisions: 4116 -> 1627; -60.5 % | Helped |
| Jointly optimize rotation and translation using the transformed point-cloud loss (ablation) | Points2Objects 9-DoF bounding-box estimation · ShapeNet-triplets validation | 3D detection mAP @ 0.5: 46.8 -> 48.6 | Helped |
| Project the predicted rotation matrix into SO(3) (ablation) | Points2Objects 9-DoF bounding-box estimation · ShapeNet-triplets validation | 3D detection mAP @ 0.5: 44.8 -> 46.8 | Helped |
| Replace rotation-bin classification and offset regression with direct rotation-matrix prediction (ablation) | Points2Objects 9-DoF bounding-box estimation · ShapeNet-triplets validation | 3D detection mAP @ 0.5: 43.3 -> 48.6 | Helped |
| Use Points2Objects rather than CoReNet m9 | Multi-object reconstruction · ShapeNet-triplets | Abs. 3D IoU mean: 43.9 -> 36.4 | Hurt |
| Use Points2Objects rather than CoReNet m9 | Multi-object reconstruction · ShapeNet-triplets | Rel. 3D IoU mean: 43.9 -> 59.5 | Helped |
| Use Points2Objects rather than CoReNet | 3D reconstruction on real images · Pix3D S1 | mIoU: 33.3% -> 34.1% | Helped |
| Use Points2Objects rather than CoReNet | 3D reconstruction on real images · Pix3D S2 | mIoU: 23.6% -> 26.3% | Helped |
| Train on ShapeNet-pairs and evaluate on ShapeNet-triplets instead of training and testing on triplets | Points2Objects · ShapeNet-triplets test | 3D mIoU: 36.4 -> 32.7 | Hurt |
Closest papers by meaning
- Semantic Multi-View Stereo: Jointly Estimating Objects and Voxels (CVPR 2017 · similarity 0.89)
- FroDO: From Detections to 3D Objects (CVPR 2020 · similarity 0.88)
- In-Hand 3D Object Reconstruction from a Monocular RGB Video (AAAI 2024 · similarity 0.88)
- Hand-held Object Reconstruction from RGB Video with Dynamic Interaction (CVPR 2025 · similarity 0.87)
- LASA: Instance Reconstruction from Real Scans using A Large-scale Aligned Shape Annotation Dataset (CVPR 2024 · similarity 0.87)
- PyTorchGeoNodes: Enabling Differentiable Shape Programs for 3D Shape Reconstruction (CVPR 2025 · similarity 0.86)
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.