Research map / 3D vision / Face and shape
High-Fidelity 3D GAN Inversion by Pseudo-Multi-View Optimization
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 |
|---|---|---|---|
| Use pseudo-multi-view optimization with visibility-aware texture warping and generated occluded regions instead of optimizing only against the input view | EG3D inversion · CelebA-HQ test | PSNR: 26.64 -> 29.43 | Helped |
| Use pseudo-multi-view optimization with visibility-aware texture warping and generated occluded regions instead of optimizing only against the input view | EG3D inversion · CelebA-HQ test | LPIPS: 0.271 -> 0.172 | Helped |
| Use pseudo-multi-view optimization with visibility-aware texture warping and generated occluded regions instead of optimizing only against the input view | EG3D inversion · CelebA-HQ test novel views | 3D consistency PSNR: 21.20 -> 21.69 | Helped |
| Add generated priors to reconstruct occluded parts instead of bilinear interpolation (ablation) | Pseudo-multi-view optimization · Novel-view face shape and boundary | visual quality | Helped |
| Add original input textures for visible parts instead of using generated textures (ablation) | Pseudo-multi-view optimization · Novel-view image-specific details | detail preservation | Helped |
| Use pseudo-multi-view supervision instead of density regularization alone (ablation) | EG3D inversion · Novel-view synthesis | visual detail quality | Helped |
| Continue single-view optimization without pseudo-view regularization | 3D GAN inversion · Input-view reconstruction versus novel-view synthesis | 3D consistency | Hurt |
| Use density regularization while retaining single-input reconstruction supervision (ablation) | EG3D inversion · Novel-view synthesis | visual detail quality | Hurt |
| Use pseudo-multi-view optimization rather than density regularization (ablation) | EG3D inversion · Novel-view appearance | visual texture quality | Helped |
| Use pseudo-multi-view optimization instead of density regularization (ablation) | EG3D inversion · Novel-view reconstruction | geometry quality | Mixed |
Closest papers by meaning
- 3D GAN Inversion With Facial Symmetry Prior (CVPR 2023 · similarity 0.93)
- WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting (NeurIPS 2025 · similarity 0.92)
- High-Fidelity GAN Inversion for Image Attribute Editing (CVPR 2022 · similarity 0.91)
- Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images (NeurIPS 2024 · similarity 0.91)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.90)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.89)
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.