Research map / 3D vision / Face and shape
One-Shot Feed-Forward 360° Animatable Avatar via Inpainted UV-Space Gaussian Modeling
Shuling Zhao, Dan Xu
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 UV shape refinement (UV offset UNet) (ablation) | Ours · HDTF self-reenactment | PSNR: 26.15 -> 26.61 | Improved |
| add local UV feature maps (ablation) | Ours · HDTF self-reenactment | CSIM: 0.5230 -> 0.8622 | Improved |
| add local UV feature maps (ablation) | Ours · MEAD self-reenactment | CSIM: 0.3932 -> 0.6765 | Improved |
| add transformer-based fusion in UV inpainting (ablation) | Ours · HDTF self-reenactment | PSNR: 25.79 -> 26.61 | Improved |
| add symmetric operations in UV inpainting (ablation) | Ours · HDTF self-reenactment | PSNR: 26.60 -> 26.61 | No clear effect |
| add symmetric operations in UV inpainting (ablation) | Ours · MEAD self-reenactment | APD: 2.2145 -> 1.9484 | Improved |
| use 3D TV loss instead of UV TV loss (ablation) | Ours · HDTF self-reenactment | PSNR: 26.40 -> 26.61 | Improved |
| add global UV feature maps from 3D GAN (ablation) | Ours · MEAD self-reenactment | APD: 2.1511 -> 1.9484 | Improved |
| add real multi-view NeRSemble training data | Ours · MEAD self-reenactment | PSNR: 17.20 -> 17.51 | Improved |
| train without pseudo multi-view data (ablation) | Ours · VFHQ self-reenactment | PSNR: 23.24 -> 23.62 | Improved |
| add lightweight expression-aware motion module | Ours · VFHQ self-reenactment | CSIM: 0.8012 -> 0.8138 | Improved |
| add pseudo multi-view supervision from 3D GAN inversion (ablation) | Ours · MEAD 360-degree view synthesis | 360 view synthesis | Improved |
Closest papers by meaning
- Generalizable and Animatable Gaussian Head Avatar (NeurIPS 2024 · similarity 0.91)
- HRAvatar: High-Quality and Relightable Gaussian Head Avatar (CVPR 2025 · similarity 0.90)
- GAF: Gaussian Avatar Reconstruction from Monocular Videos via Multi-view Diffusion (CVPR 2025 · similarity 0.90)
- RGBAvatar: Reduced Gaussian Blendshapes for Online Modeling of Head Avatars (CVPR 2025 · similarity 0.90)
- FATE: Full-head Gaussian Avatar with Textural Editing from Monocular Video (CVPR 2025 · similarity 0.90)
- Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians (CVPR 2024 · similarity 0.90)
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.
