Research map / 3D vision / 3D generation
GAS: Generative Avatar Synthesis from a Single Image
Yixing Lu, Junting Dong, Youngjoong Kwon, Qin Zhao, Bo Dai, Fernando De la Torre
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 SMPL normal-map geometry cue (ablation) | GAS · THuman novel view synthesis | PSNR: 26.07 -> 26.77 | Improved |
| add SMPL normal-map geometry cue (ablation) | GAS · MVHumanNet novel pose synthesis | PSNR: 28.33 -> 28.74 | Improved |
| add SMPL normal-map geometry cue (ablation) | GAS · THuman novel view synthesis | FVD: 227.1 -> 194.8 | Improved |
| add NeRF appearance cue (ablation) | GAS · THuman novel view synthesis | PSNR: 26.38 -> 26.77 | Improved |
| add NeRF appearance cue (ablation) | GAS · MVHumanNet novel pose synthesis | PSNR: 27.66 -> 28.74 | Improved |
| add NeRF appearance cue (ablation) | GAS · MVHumanNet novel pose synthesis | FVD: 234.6 -> 188.5 | Improved |
| add video diffusion refinement on NeRF renderings (ablation) | GAS · 2K2K novel view synthesis | PSNR: 24.25 -> 28.82 | Improved |
| add video diffusion refinement on NeRF renderings (ablation) | GAS · TikTok-style pose synthesis (MVHumanNet pose) | FVD: 1255.8 -> 362.0 | Improved |
| add task switcher (ablation) | GAS · THuman novel view synthesis | PSNR: 26.58 -> 26.77 | Improved |
| add task switcher (ablation) | GAS · THuman novel view synthesis | FVD: 198.9 -> 194.8 | Improved |
| add dynamic videos to 3D scan training for novel view synthesis (ablation) | GAS · THuman novel view synthesis | PSNR: 26.82 -> 26.77 | No clear effect |
| add dynamic videos to 3D scan training for novel view synthesis (ablation) | GAS · 2K2K novel view synthesis | FVD: 187.9 -> 191.3 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Champ — Champ: Controllable and Consistent Human Image Animation with 3D Parametric Guidance (ECCV 2024) · better on 6 of 6 shared results
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
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.92)
- Morphable Diffusion: 3D-Consistent Diffusion for Single-image Avatar Creation (CVPR 2024 · similarity 0.91)
- En3D: An Enhanced Generative Model for Sculpting 3D Humans from 2D Synthetic Data (CVPR 2024 · similarity 0.90)
- GAF: Gaussian Avatar Reconstruction from Monocular Videos via Multi-view Diffusion (CVPR 2025 · similarity 0.90)
- Zero-1-to-A: Zero-Shot One Image to Animatable Head Avatars Using Video Diffusion (CVPR 2025 · similarity 0.90)
- MagicMan: Generative Novel View Synthesis of Humans with 3D-Aware Diffusion and Iterative Refinement (AAAI 2025 · similarity 0.89)
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.
