Research map / 3D vision / Face and shape
Avat3r: Large Animatable Gaussian Reconstruction Model for High-fidelity 3D Head Avatars
Tobias Kirschstein, Javier Romero, Artem Sevastopolsky, Matthias Nießner, Shunsuke Saito
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 Sapiens feature maps (ablation) | Avat3r · Ava256 ablation (no LPIPS loss, 667x667) | PSNR: 20.9 -> 21.6 | Improved |
| add DUSt3R position maps (ablation) | Avat3r · Ava256 ablation (no LPIPS loss, 667x667) | AKD: 9.60 -> 8.08 | Improved |
| add DUSt3R position maps (ablation) | Avat3r · Ava256 ablation (no LPIPS loss, 667x667) | PSNR: 21.1 -> 21.6 | Improved |
| add random input timesteps in training (ablation) | Avat3r · Ava256 ablation (no LPIPS loss, 667x667) | AKD: 8.86 -> 8.08 | Improved |
| add random input timesteps in training (ablation) | Avat3r · Ava256 ablation (no LPIPS loss, 667x667) | LPIPS: 0.409 -> 0.410 | No clear effect |
| add both Sapiens and DUSt3R over Dynamic GRM (ablation) | Dynamic GRM · Ava256 ablation (no LPIPS loss, 667x667) | PSNR: 20.8 -> 21.6 | Improved |
| add both skip connections (ablation) | Avat3r · Skip connection ablation (no LPIPS loss, 667x667) | PSNR: 21.39 -> 22.05 | Improved |
| add position skip connection (ablation) | Avat3r · Skip connection ablation (no LPIPS loss, 667x667) | PSNR: 21.55 -> 22.05 | Improved |
| add color skip connection (ablation) | Avat3r · Skip connection ablation (no LPIPS loss, 667x667) | PSNR: 21.76 -> 22.05 | Improved |
| add position skip connection (ablation) | Avat3r · Skip connection ablation (no LPIPS loss, 667x667) | AKD: 7.69 -> 7.99 | Hurt |
| fine-tune on 984 additional neutral identities | Avat3r · NeRSemble (4 input views) | CSIM: 0.50 -> 0.76 | Improved |
| compare Avat3r to GPAvatar | GPAvatar · Ava256 (4 input views) | PSNR: 19.4 -> 20.7 | Improved |
Later papers that compared with this method
4 later accepted papers report Avat3r as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- FastGHA: Generalized Few-Shot 3D Gaussian Head Avatars with Real-Time Animation (ICLR 2026 · better on 4 of 5 shared results)
- FlexAvatar: Learning Complete 3D Head Avatars with Partial Supervision (CVPR 2026 · better on 3 of 3 shared results)
- Feed-forward Gaussian Registration for Head Avatar Creation and Editing (CVPR 2026 · better on 1 of 1 shared result)
- Large-Scale High-Quality 3D Gaussian Head Reconstruction from Multi-View Captures (ECCV 2026 · better on 1 of 1 shared result)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GPAvatar — GPAvatar: High-fidelity Head Avatars by Learning Efficient Gaussian Projections (CVPR 2025) · better on 6 of 6 shared results
- HeadNeRF — HeadNeRF: A Real-Time NeRF-Based Parametric Head Model (CVPR 2022) · 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
- GAF: Gaussian Avatar Reconstruction from Monocular Videos via Multi-view Diffusion (CVPR 2025 · similarity 0.90)
- Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians (CVPR 2024 · similarity 0.89)
- FA-LAM: Focus-Aware Large Avatar Model for One-Shot 4D Animatable Gaussian Head (NeurIPS 2026 · similarity 0.89)
- Generalizable and Animatable Gaussian Head Avatar (NeurIPS 2024 · similarity 0.89)
- HRAvatar: High-Quality and Relightable Gaussian Head Avatar (CVPR 2025 · similarity 0.89)
- RGBAvatar: Reduced Gaussian Blendshapes for Online Modeling of Head Avatars (CVPR 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.
