Research map / 3D vision / Face and shape
FaceLit: Neural 3D Relightable Faces
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 FaceLit-D instead of EG3D 512² | FFHQ generation · FFHQ | FID: 4.70 -> 3.48 | Helped |
| Use FaceLit-F instead of EG3D 512² | FFHQ generation · FFHQ | Pose: 0.005 -> 0.0008 | Helped |
| Add the specular component to the 64² diffuse-only model (ablation) | FaceLit on FFHQ · FFHQ | Depth: 0.42 -> 0.33 | Helped |
| Add the specular component to the 128² diffuse-only model (ablation) | FaceLit on FFHQ · FFHQ | Depth: 0.62 -> 0.43 | Helped |
| Add the specular component to the 128² diffuse-only model (ablation) | FaceLit on FFHQ · FFHQ | FID: 3.48 -> 3.90 | Hurt |
| Increase rendering resolution from 64² to 128² in the diffuse-only model (ablation) | FaceLit on FFHQ · FFHQ | FID: 4.01 -> 3.48 | Helped |
| Increase rendering resolution from 64² to 128² in the diffuse-only model (ablation) | FaceLit on FFHQ · FFHQ | Depth: 0.42 -> 0.62 | Hurt |
| Add the specular component to the diffuse-only model (ablation) | FaceLit on MetFaces · MetFaces | Depth: 0.96 -> 0.77 | Helped |
| Add the specular component to the diffuse-only model (ablation) | FaceLit on MetFaces · MetFaces | FID: 15.30 -> 15.43 | Hurt |
| Add the specular component to the diffuse-only model (ablation) | FaceLit on CelebA-HQ · CelebA-HQ | Depth: 0.36 -> 0.33 | Helped |
| Add the specular component to the diffuse-only model (ablation) | FaceLit on CelebA-HQ · CelebA-HQ | FID: 3.63 -> 3.94 | Hurt |
| Use the full model rather than diffuse-only (ablation) | FaceLit on FFHQ · FFHQ | Illumination SH MSE: 0.0054 -> 0.0053 | Helped |
Closest papers by meaning
- Next3D: Generative Neural Texture Rasterization for 3D-Aware Head Avatars (CVPR 2023 · similarity 0.88)
- Lifting 2D StyleGAN for 3D-Aware Face Generation (CVPR 2021 · similarity 0.88)
- MOST-GAN: 3D Morphable StyleGAN for Disentangled Face Image Manipulation (AAAI 2022 · similarity 0.88)
- In-N-Out: Faithful 3D GAN Inversion with Volumetric Decomposition for Face Editing (CVPR 2024 · similarity 0.88)
- BecomingLit: Relightable Gaussian Avatars with Hybrid Neural Shading (NeurIPS 2025 · similarity 0.87)
- Controllable 3D Face Generation with Conditional Style Code Diffusion (AAAI 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.