Research map / 3D vision / Face and shape
Generative Multiplane Images: Making a 2D GAN 3D-Aware
Xiaoming Zhao, Fangchang Ma, David Güera, Zhile Ren, Alexander G. Schwing, Alex Colburn
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 discriminator pose conditioning to the MPI branch (ablation) | GMPI (naive MPI, 32 planes) · FFHQ 512 | Depth error: 2.190 -> 0.501 | Improved |
| add discriminator pose conditioning to the MPI branch (ablation) | GMPI (naive MPI, 32 planes) · FFHQ 512 | Pose error: 0.062 -> 0.006 | Improved |
| add plane-specific depth-conditioned feature without pose conditioning (ablation) | GMPI · FFHQ 512 | Depth error: 2.190 -> 2.140 | No clear effect |
| add plane-specific feature on top of discriminator pose conditioning (ablation) | GMPI with DPC · FFHQ 512 | Depth error: 0.501 -> 0.462 | Improved |
| add discriminator pose conditioning (3D-awareness) at cost of 2D quality (ablation) | GMPI with plane feature · FFHQ 512 | FID: 4.35 -> 7.19 | Hurt |
| add shading-guided training (ablation) | GMPI with DPC and plane feature · FFHQ 512 | Depth error: 0.462 -> 0.457 | No clear effect |
| increase planes at inference from 32 to 96 (ablation) | GMPI · FFHQ 512 | FID: 7.40 -> 8.29 | Hurt |
| increase training planes from 4 to 32 (ablation) | GMPI · FFHQ 512 | Depth error: 0.992 -> 0.457 | Improved |
| increase training planes from 4 to 32 (ablation) | GMPI · FFHQ 512 | FID: 16.4 -> 7.40 | Improved |
| add camera pose noise during training (0 to 4 sigma) (ablation) | GMPI · FFHQ 512 | Depth error: 0.46 -> 1.68 | Hurt |
| lower truncation psi from 1.0 to 0.5 at inference (ablation) | GMPI · FFHQ 256 | Depth error: 0.533 -> 0.445 | Improved |
| predict alpha maps directly instead of from a depth map (D2A) (ablation) | GMPI · FFHQ 256 | Depth error: 0.63 -> 0.53 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 5 on most shared results.
- StyleNeRF — StyleNeRF: A Style-based 3D Aware Generator for High-resolution Image Synthesis (ICLR 2022) · better on 2 of 4 shared results
- GRAM — GRAM: Generative Radiance Manifolds for 3D-Aware Image Generation (CVPR 2022) · better on 1 of 1 shared result
- StyleSDF — StyleSDF: High-Resolution 3D-Consistent Image and Geometry Generation (CVPR 2022) · better on 1 of 1 shared result
- GIRAFFE — GIRAFFE: Representing Scenes As Compositional Generative Neural Feature Fields (CVPR 2021) · better on 1 of 1 shared result
- pi-GAN 128 — Pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesis (CVPR 2021) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Dual Mapping of 2D StyleGAN for 3D-Aware Image Generation and Manipulation (Student Abstract) (AAAI 2024 · similarity 0.90)
- Efficient Geometry-Aware 3D Generative Adversarial Networks (CVPR 2022 · similarity 0.90)
- EpiGRAF: Rethinking training of 3D GANs (NeurIPS 2022 · similarity 0.89)
- Injecting 3D Perception of Controllable NeRF-GAN into StyleGAN for Editable Portrait Image Synthesis (ECCV 2022 · similarity 0.89)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.88)
- Reference-Based 3D-Aware Image Editing with Triplanes (CVPR 2025 · 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.
