Research map / 3D vision / Face and shape
Im2Struct: Recovering 3D Shape Structure From a Single RGB Image
Read the paperCheck 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 |
|---|---|---|---|
| Add the two-scale refinement network to the single-scale masking network (ablation) | Structure masking network · Test set object masks | overall pixel accuracy: 0.953 -> 0.988 | Helped |
| Add the two-scale refinement network to the single-scale masking network (ablation) | Structure masking network · Test set object masks | per-class accuracy: 0.917 -> 0.983 | Helped |
| Add jump connections (ablation) | Two-scale structure masking network · Test set object masks | overall pixel accuracy: 0.982 -> 0.988 | Helped |
| Add jump connections (ablation) | Two-scale structure masking network · Test set object masks | per-class accuracy: 0.964 -> 0.983 | Helped |
| Replace vanilla VGG-16 features with structure masking features (ablation) | 3D structure recovery · Test set 3D shape structures | Hausdorff error: 0.0980 -> 0.0894 | Helped |
| Replace vanilla VGG-16 features with structure masking features (ablation) | 3D structure recovery · Test set 3D shape structures | thresholded accuracy (δ<0.1): 67.8% -> 75.3% | Helped |
| Replace vanilla VGG-19 features with structure masking features (ablation) | 3D structure recovery · Test set 3D shape structures | Hausdorff error: 0.0922 -> 0.0846 | Helped |
| Replace vanilla VGG-19 features with structure masking features (ablation) | 3D structure recovery · Test set 3D shape structures | thresholded accuracy (δ<0.1): 72.2% -> 78.5% | Helped |
| Use VGG-19 rather than VGG-16 in the structure masking network (ablation) | Structure masking network · Test set 3D shape structures | Hausdorff error: 0.0894 -> 0.0846 | Helped |
| Use VGG-19 rather than VGG-16 in the structure masking network (ablation) | Structure masking network · Test set 3D shape structures | thresholded accuracy (δ<0.2): 97.8% -> 97.6% | No clear effect |
| Use VGG-19 rather than VGG-16 in the structure masking network (ablation) | Structure masking network · Test set 3D shape structures | thresholded accuracy (δ<0.1): 75.3% -> 78.5% | Helped |
Closest papers by meaning
- Learning Unsupervised Hierarchical Part Decomposition of 3D Objects From a Single RGB Image (CVPR 2020 · similarity 0.92)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.90)
- RIM-Net: Recursive Implicit Fields for Unsupervised Learning of Hierarchical Shape Structures (CVPR 2022 · similarity 0.89)
- Cuboids Revisited: Learning Robust 3D Shape Fitting to Single RGB Images (CVPR 2021 · similarity 0.89)
- Neural Parts: Learning Expressive 3D Shape Abstractions With Invertible Neural Networks (CVPR 2021 · similarity 0.89)
- 3DCoMPaT200: Language Grounded Large-Scale 3D Vision Dataset for Compositional Recognition (NeurIPS 2024 · 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.