Research map / 3D vision / Face and shape
CoReNet: Coherent 3D Scene Reconstruction from a Single RGB Image
Stefan Popov, Pablo Bauszat, Vittorio Ferrari
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 ray-traced skip connections (focal loss) (ablation) | CoReNet · ShapeNet single object, 13 classes, low realism | mIoU (%): 50.8 -> 56.6 | Improved |
| add ray-traced skip connections (IoU loss) (ablation) | CoReNet · ShapeNet single object, 13 classes, low realism | mIoU (%): 53.0 -> 57.9 | Improved |
| add ray-traced skip connections (IoU loss) (ablation) | CoReNet · ShapeNet single object, 13 classes, low realism | F@1% (%): 47.8 -> 57.5 | Improved |
| use IoU loss instead of Xent (ablation) | CoReNet with skip connections · ShapeNet single object, 13 classes, low realism | mIoU (%): 54.1 -> 57.9 | Improved |
| use IoU loss instead of focal loss (ablation) | CoReNet with skip connections · ShapeNet single object, 13 classes, low realism | mIoU (%): 56.6 -> 57.9 | Improved |
| render with high instead of low image realism (ablation) | CoReNet · ShapeNet single object, 13 classes | mIoU (%): 57.9 -> 59.1 | Improved |
| add ray-traced skip connections (focal loss) (ablation) | CoReNet · ShapeNet multi-object pairs, high realism | mIoU (%): 34.9 -> 42.7 | Improved |
| add ray-traced skip connections (IoU loss) (ablation) | CoReNet · ShapeNet multi-object pairs, high realism | mIoU (%): 33.1 -> 43.1 | Improved |
| use IoU loss instead of Xent (ablation) | CoReNet with skip connections · ShapeNet multi-object pairs, high realism | mIoU (%): 30.0 -> 43.1 | Improved |
| use IoU loss instead of focal loss (ablation) | CoReNet with skip connections · ShapeNet multi-object pairs, high realism | mIoU (%): 42.7 -> 43.1 | No clear effect |
| train at 64^3 grid and sample 8 offsets to reach 128^3 instead of training at 128^3 (ablation) | CoReNet hybrid representation · ShapeNet single object, low realism, test at 128^3 from lower-res training | mIoU (%): 56.6 -> 56.1 | No clear effect |
| train at 32^3 grid and sample 64 offsets to reach 128^3 instead of training at 128^3 (ablation) | CoReNet hybrid representation · ShapeNet single object, low realism, test at 128^3 from lower-res training | mIoU (%): 56.6 -> 53.1 | Hurt |
Later papers that compared with this method
2 later accepted papers report CoReNet as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- Voxel-based 3D Detection and Reconstruction of Multiple Objects from a Single Image (NeurIPS 2021 · better on 7 of 7 shared results)
- From Points to Multi-Object 3D Reconstruction (CVPR 2021 · better on 2 of 3 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
- Coherent 3D Scene Diffusion From a Single RGB Image (NeurIPS 2024 · similarity 0.90)
- 3D Scene Reconstruction from a Single Viewport (ECCV 2020 · similarity 0.90)
- What Do Single-View 3D Reconstruction Networks Learn? (CVPR 2019 · similarity 0.89)
- Single-View 3D Object Reconstruction From Shape Priors in Memory (CVPR 2021 · similarity 0.89)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.88)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · 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.
