Research map / 3D vision / 3D generation
Make Encoder Great Again in 3D GAN Inversion through Geometry and Occlusion-Aware Encoding
Ziyang Yuan, Yiming Zhu, Yu Li, Hongyu Liu, Chun Yuan
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 canonical latent discriminator and background depth regularization (ablation) | geometry-aware encoder · MEAD yaw=+-60 | Geo. Err.: 0.1422 -> 0.0943 | Improved |
| add canonical latent discriminator only (without background regularization) (ablation) | geometry-aware encoder · MEAD yaw=+-60 | Geo. Err.: 0.1422 -> 0.1414 | No clear effect |
| add background regularization only (without discriminator) (ablation) | geometry-aware encoder · MEAD yaw=+-60 | Geo. Err.: 0.1422 -> 0.1016 | Improved |
| replace CNN with adaptive feature alignment (cross-attention) module (ablation) | feature refinement · CelebA-HQ | FID: 25.1 -> 18.1 | Improved |
| add occlusion-aware mix tri-plane (ablation) | full method · MEAD yaw=+-60 | LPIPS: 0.2327 -> 0.2192 | Improved |
| add occlusion-aware mix tri-plane (ablation) | full method · MEAD yaw=0 | MSE: 0.03202 -> 0.0329 | Mixed |
| use encoder with AFA instead of optimization (PTI) | 3D GAN inversion · CelebA-HQ input view | inference time (s): 115.2 -> 0.2262 | Improved |
| use encoder-based inversion instead of optimization at extreme yaw | 3D GAN inversion · MEAD yaw=+-60 | FID: 69.8 -> 50.9 | Improved |
| add AFA feature refinement on top of w+ encoder | Ours-w+ · CelebA-HQ input view | MSE: 0.0697 -> 0.0301 | Improved |
Closest papers by meaning
- $$E^{3}DGE$$: Self-Supervised Geometry-Aware Encoder for Style-Based 3D GAN Inversion (IJCV 2025 · similarity 0.92)
- Dual Encoder GAN Inversion for High-Fidelity 3D Head Reconstruction from Single Images (NeurIPS 2024 · similarity 0.92)
- Self-Supervised Geometry-Aware Encoder for Style-Based 3D GAN Inversion (CVPR 2023 · similarity 0.91)
- WarpGAN: Warping-Guided 3D GAN Inversion with Style-Based Novel View Inpainting (NeurIPS 2025 · similarity 0.90)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.88)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 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.
