Research map / 3D vision / Face and shape
Learning to Reconstruct High-quality 3D Shapes with Cascaded Fully Convolutional Networks
Yan-Pei Cao, Zheng-Ning Liu, Zheng-Fei Kuang, Leif Kobbelt, Shi-Min Hu
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 visual hull input branches (ablation) | 3D-CFCN (2 views) · Human10 reconstruction | relative Hausdorff RMS distance: 0.0035 -> 0.0031 | Improved |
| add visual hull input branches (ablation) | 3D-CFCN (2 views) · ModelNet40 reconstruction | relative Hausdorff RMS distance: 0.0032 -> 0.0019 | Improved |
| fuse 4 viewpoints instead of 2 (ablation) | 3D-CFCN · Human10 reconstruction | relative Hausdorff RMS distance: 0.0035 -> 0.0021 | Improved |
| fuse 4 viewpoints instead of 2 (ablation) | 3D-CFCN · ModelNet40 reconstruction | relative Hausdorff RMS distance: 0.0032 -> 0.0010 | Improved |
| use 3D-CFCN instead of OctNetFusion | 3D-CFCN vs OctNetFusion · Human10 reconstruction | relative Hausdorff RMS distance: 0.0040 -> 0.0035 | Improved |
| use 3D-CFCN instead of OctNetFusion | 3D-CFCN vs OctNetFusion · ModelNet40 reconstruction | relative Hausdorff RMS distance: 0.0035 -> 0.0032 | No clear effect |
| use 3D-CFCN instead of 3D-EPN | 3D-CFCN vs 3D-EPN · Human10 reconstruction | relative Hausdorff RMS distance: 0.0263 -> 0.0035 | Improved |
| use 3D-CFCN instead of PSR | 3D-CFCN vs PSR · ModelNet40 reconstruction | relative Hausdorff RMS distance: 0.0620 -> 0.0032 | Improved |
| add second refinement stage (ablation) | 3D-CFCN · Human10 reconstruction | visual quality | Improved |
| add second refinement stage (ablation) | 3D-CFCN · ModelNet40 reconstruction | visual quality | No clear effect |
| use 3D-CFCN decoder instead of OGN decoder on same features | 3D-CFCN vs OGN · Human10 occupancy maps (512^3) | visual fidelity | Improved |
| perturb camera poses by 2.5% to 10% at test time | 3D-CFCN · calibration error test | visual quality | Mixed |
Closest papers by meaning
- StereoDRNet: Dilated Residual StereoNet (CVPR 2019 · similarity 0.89)
- Image-Based 3D Object Reconstruction: State-of-the-Art and Trends in the Deep Learning Era (TPAMI 2019 · similarity 0.89)
- High-Resolution Shape Completion Using Deep Neural Networks for Global Structure and Local Geometry Inference (ICCV 2017 · similarity 0.89)
- Learning Non-Volumetric Depth Fusion Using Successive Reprojections (CVPR 2019 · similarity 0.88)
- Dense 3D Object Reconstruction from a Single Depth View (TPAMI 2018 · similarity 0.88)
- Implicit Functions in Feature Space for 3D Shape Reconstruction and Completion (CVPR 2020 · 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.
