Research map / 3D vision / Gaussian splatting
Large-Scale High-Quality 3D Gaussian Head Reconstruction from Multi-View Captures
Evan Ntavelis, Sean Wu, Mohamad Shahbazi, Fabio Maninchedda, Dmitry Kostiaev, Artem Sevastopolsky, Mehak Gupta, Vittorio Megaro, Trevor Phillips, Thomas Etterlin, Jeronimo Bayer, Simon Schaefer and 12 more
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 |
|---|---|---|---|
| use HeadsUp UV-Gaussian feed-forward model instead of Avat3r | HeadsUp · Internal10K | PSNR: 24.37 -> 29.25 | Improved |
| use HeadsUp instead of Avat3r | HeadsUp · Internal10K | SSIM: 0.831 -> 0.821 | Hurt |
| use HeadsUp instead of Avat3r | HeadsUp · Ava-256 | PSNR: 22.31 -> 24.62 (6 views) | Improved |
| use fixed neutral template instead of expression-tracked mesh (ablation) | HeadsUp · Internal10K ablation | PSNR: 26.77 -> 28.89 | Improved |
| add vertex position loss (ablation) | HeadsUp · Internal10K ablation | PSNR: 27.68 -> 28.46 | Improved |
| scale training subjects from 250 to 10K (ablation) | HeadsUp · Internal10K ablation | PSNR: 22.19 -> 28.89 | Improved |
| increase input views from 1 to 16 (ablation) | HeadsUp · Internal10K ablation | PSNR: 22.98 -> 29.03 | Improved |
| increase latent size from 16 to 128 (ablation) | HeadsUp · Internal10K ablation | PSNR: 27.59 -> 29.52 (UV 256) | Improved |
| double Gaussian UV resolution 256 to 512 (ablation) | HeadsUp · Internal10K ablation (latent 32) | PSNR: 28.89 -> 28.90 | No clear effect |
| supervise with 8 target views instead of 1 (ablation) | HeadsUp · Internal10K ablation | PSNR: 28.89 -> 29.54 | Improved |
| use 12 encoder blocks instead of 8 (ablation) | HeadsUp · Internal10K ablation | PSNR: 28.89 -> 28.89 | No clear effect |
| add eye and mouth region losses in finetuning (ablation) | HeadsUp · high-res finetuning | eye/mouth crop PSNR: -1.56 dB eye, -1.97 dB mouth when removed | Improved |
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.
- FastGHA — FastGHA: Generalized Few-Shot 3D Gaussian Head Avatars with Real-Time Animation (ICLR 2026) · better on 1 of 1 shared result
- Avat3r — Avat3r: Large Animatable Gaussian Reconstruction Model for High-fidelity 3D Head Avatars (ICCV 2025) · better on 1 of 1 shared result
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
- Multi-view Consistent 3D Gaussian Head Avatars 'without' Multi-view Generation (CVPR 2026 · similarity 0.91)
- Gaussian Head Avatar: Ultra High-fidelity Head Avatar via Dynamic Gaussians (CVPR 2024 · similarity 0.89)
- CGS-GAN: 3D Consistent Gaussian Splatting GANs for High Resolution Human Head Synthesis (NeurIPS 2025 · similarity 0.88)
- FA-LAM: Focus-Aware Large Avatar Model for One-Shot 4D Animatable Gaussian Head (NeurIPS 2026 · similarity 0.88)
- GAF: Gaussian Avatar Reconstruction from Monocular Videos via Multi-view Diffusion (CVPR 2025 · similarity 0.88)
- Gaussian Head & Shoulders: High Fidelity Neural Upper Body Avatars with Anchor Gaussian Guided Texture Warping (ICLR 2025 · 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.
