Research map / 3D vision / Face and shape
EpiGRAF: Rethinking training of 3D GANs
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 |
|---|---|---|---|
| replace annealed uniform patch-scale sampling with beta sampling (T = 5k) (ablation) | GRAF with tri-planes · FFHQ 512² | FID: 13.41 -> 11.57 | Helped |
| replace annealed uniform patch-scale sampling with beta sampling (T = 5k) (ablation) | GRAF with tri-planes · M-Plants 256² | FID: 24.99 -> 21.77 | Helped |
| use one scale/location-aware discriminator rather than the standard discriminator (ablation) | EpiGRAF · FFHQ 512² | FID: 13.41 -> 9.92 | Helped |
| use one scale/location-aware discriminator rather than the standard discriminator (ablation) | EpiGRAF · M-Plants 256² | FID: 24.99 -> 19.42 | Helped |
| use 2 scale-specific discriminators instead of one scale/location-aware discriminator (ablation) | EpiGRAF · FFHQ 512² | FID: 9.92 -> 10.87 | Hurt |
| use 2 scale-specific discriminators instead of one scale/location-aware discriminator (ablation) | EpiGRAF · M-Plants 256² | FID: 19.42 -> 21.02 | Hurt |
| use 4 scale-specific discriminators instead of one scale/location-aware discriminator (ablation) | EpiGRAF · FFHQ 512² | FID: 9.92 -> 21.56 | Hurt |
| use 4 scale-specific discriminators instead of one scale/location-aware discriminator (ablation) | EpiGRAF · M-Plants 256² | FID: 19.42 -> 43.11 | Hurt |
| reduce patch resolution from 64² to 32² (ablation) | EpiGRAF · FFHQ 512² | FID: 9.92 -> 17.44 | Hurt |
| reduce patch resolution from 64² to 32² (ablation) | EpiGRAF · M-Plants 256² | FID: 19.42 -> 34.32 | Hurt |
| increase patch resolution from 64² to 128² (ablation) | EpiGRAF · FFHQ 512² | FID: 9.92 -> 11.36 | Hurt |
| increase patch resolution from 64² to 128² (ablation) | EpiGRAF · M-Plants 256² | FID: 19.42 -> 18.90 | Helped |
Closest papers by meaning
- 3D generation on ImageNet (ICLR 2023 · similarity 0.92)
- What You See is What You GAN: Rendering Every Pixel for High-Fidelity Geometry in 3D GANs (CVPR 2024 · similarity 0.91)
- Efficient Geometry-Aware 3D Generative Adversarial Networks (CVPR 2022 · similarity 0.91)
- CAD: Photorealistic 3D Generation via Adversarial Distillation (CVPR 2024 · similarity 0.89)
- G-NeRF: Geometry-enhanced Novel View Synthesis from Single-View Images (CVPR 2024 · similarity 0.89)
- WildFusion: Learning 3D-Aware Latent Diffusion Models in View Space (ICLR 2024 · similarity 0.88)
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.