Research map / 3D vision / Face and shape
Unsupervised 3D Shape Completion Through GAN Inversion
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 |
|---|---|---|---|
| add PatchVariance uniformity loss during tree-GAN training (ablation) | tree-GAN · ShapeNet shape generation, Average | MMD-EMD (×10^3): 47.7 -> 39.0 | Helped |
| replace k-Mask with voxel-Mask (ablation) | ShapeInversion · ShapeNet virtual scan benchmark | CD (×10^4): 14.9 -> 19.3 | Hurt |
| replace k-Mask with τ-Mask (ablation) | ShapeInversion · ShapeNet virtual scan benchmark | CD (×10^4): 14.9 -> 18.9 | Hurt |
| add k-Mask instead of voxel-Mask (ablation) | ShapeInversion · ShapeNet virtual scan benchmark | F1: 81.5 -> 83.9 | Helped |
| add Feature Distance loss (ablation) | ShapeInversion · ShapeNet virtual scan benchmark | F1: 81.9 -> 83.9 | Helped |
| add UHD loss to the inversion objective (ablation) | ShapeInversion · KITTI Car real scans | UHD (×10^2): 13.8 -> 12.5 | Helped |
| add UHD loss to the inversion objective (ablation) | ShapeInversion · KITTI Car real scans | UCD (×10^4): 2.9 -> 5.3 | Hurt |
| use GAN inversion with a complete-shape prior instead of pcl2pcl | ShapeNet virtual scan benchmark · ShapeNet virtual scan, Average | F1: 74.2 -> 83.9 | Helped |
| use GAN inversion without real-scan training rather than pcl2pcl retrained on real scans | ShapeInversion · KITTI Car real scans | UCD (×10^4): 9.2 -> 2.9 | Helped |
Closest papers by meaning
- Diverse Shape Completion via Style Modulated Generative Adversarial Networks (NeurIPS 2023 · similarity 0.91)
- Do 2D GANs Know 3D Shape? Unsupervised 3D Shape Reconstruction from 2D Image GANs (ICLR 2021 · similarity 0.91)
- DECOR-GAN: 3D Shape Detailization by Conditional Refinement (CVPR 2021 · similarity 0.90)
- High-Fidelity 3D GAN Inversion by Pseudo-Multi-View Optimization (CVPR 2023 · similarity 0.89)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.87)
- Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images (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.