Research map / 3D vision / Face and shape
Bridging the Gap: Studio-like Avatar Creation from a Monocular Phone Capture
ShahRukh Athar, Shunsuke Saito, Stanislav Pidhorskyi, Zhengyu Yang, Chen Cao
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 |
|---|---|---|---|
| finetune StyleGAN2 by sampling inverted W+ vectors instead of Z space (ablation) | Studio-StyleGAN2 · Paired phone-studio textures (10 pairs) | LPIPS: 0.401 -> 0.364 | Improved |
| finetune with W+ sampling instead of Z space (ablation) | Studio-StyleGAN2 · Paired phone-studio textures (10 pairs) | PSNR: 21.368 -> 22.76 | Improved |
| add phone-texture gradient conditioning at both training and inference of the diffusion model (ablation) | detail diffusion model · Paired phone-studio textures (10 pairs) | DISTS: 0.179 -> 0.163 | Improved |
| add phone-texture gradient conditioning at inference only (ablation) | detail diffusion model · Paired phone-studio textures (10 pairs) | DISTS: 0.179 -> 0.183 | Hurt |
| add phone-texture gradient conditioning at inference only (ablation) | detail diffusion model · Paired phone-studio textures (10 pairs) | LPIPS: 0.383 -> 0.376 | Improved |
| use proposed StyleGAN2 W+ translation instead of CycleGAN | texture translation · Unpaired phone textures (31) | FaceID distance: 5.19e-4 -> 4.31e-4 | Improved |
| use proposed method instead of AgileGAN | texture translation · Paired phone-studio textures (10 pairs) | PSNR: 18.05 -> 22.76 | Improved |
| transform expression textures to studio-like lighting via gain/bias color transform | AVA reanimation · Reanimated avatars (qualitative) | visual quality | Improved |
Closest papers by meaning
- 3DToonify: Creating Your High-Fidelity 3D Stylized Avatar Easily from 2D Portrait Images (CVPR 2024 · similarity 0.88)
- AvatarCap: Animatable Avatar Conditioned Monocular Human Volumetric Capture (ECCV 2022 · similarity 0.86)
- Normalized Avatar Synthesis Using StyleGAN and Perceptual Refinement (CVPR 2021 · similarity 0.86)
- Codec Avatar Studio: Paired Human Captures for Complete, Driveable, and Generalizable Avatars (NeurIPS 2024 · similarity 0.86)
- BecomingLit: Relightable Gaussian Avatars with Hybrid Neural Shading (NeurIPS 2025 · similarity 0.85)
- GAF: Gaussian Avatar Reconstruction from Monocular Videos via Multi-view Diffusion (CVPR 2025 · similarity 0.85)
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.
