Research map / 3D vision / Face and shape
Generating Diverse 3D Reconstructions From a Single Occluded Face Image
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 Global+Local fitting instead of FLAME fitting | 3D face fitting on occluded CoMA images · CoMA test, overall | MSE × 10^-3: 41.24 -> 35.85 | Helped |
| Add DPP diversification to Global+Local+VAE (ablation) | Diverse3DFace · CoMA test, overall | ASD-O × 10^-3: 1.59 -> 6.05 | Helped |
| Add DPP diversification to Global+Local+VAE (ablation) | Diverse3DFace · CoMA test, overall | ASD-V × 10^-3: 0.95 -> 0.73 | Helped |
| Replace Global+Local fitting with FLAME fitting (ablation) | VAE and DPP shape completion · CoMA test, overall | CSE × 10^-3: 33.71 -> 37.45 | Hurt |
| Replace Global+Local fitting with FLAME fitting (ablation) | VAE and DPP shape completion · CoMA test, overall | ASD-O × 10^-3: 6.05 -> 5.92 | Hurt |
| Remove Mesh-VAE completion and use DPP fitting directly (ablation) | Global+Local shape fitting · CoMA test, overall | CSE × 10^-3: 33.71 -> 34.55 | Hurt |
| Remove Mesh-VAE completion and use DPP fitting directly (ablation) | Global+Local shape fitting · CoMA test, overall | ASD-O × 10^-3: 6.05 -> 3.39 | Hurt |
| Use one-pass VAE samples without DPP optimization (ablation) | Global+Local+VAE · CoMA test, overall | CSE × 10^-3: 33.71 -> 31.18 | Helped |
| Use one-pass VAE samples without DPP optimization (ablation) | Global+Local+VAE · CoMA test, overall | ASD-O × 10^-3: 6.05 -> 1.59 | Hurt |
| Increase the DPP prior-quality parameter n_sigma (ablation) | Diverse3DFace · CoMA test | ASD-O: 3.63 -> 8.64 | Helped |
Closest papers by meaning
- OFER: Occluded Face Expression Reconstruction (CVPR 2025 · similarity 0.91)
- Extreme 3D Face Reconstruction: Seeing Through Occlusions (CVPR 2018 · similarity 0.89)
- Pixel3DMM: Versatile Screen-Space Priors for Single-Image 3D Face Reconstruction (ICLR 2026 · similarity 0.88)
- Learning Complete 3D Morphable Face Models From Images and Videos (CVPR 2021 · similarity 0.87)
- Robust 3D Shape Reconstruction in Zero-Shot from a Single Image in the Wild (CVPR 2025 · similarity 0.87)
- Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image (NeurIPS 2024 · 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.