Research map / 3D vision / Face and shape
Preface: A Data-driven Volumetric Prior for Few-shot Ultra High-resolution Face Synthesis
Marcel C. Bühler, Kripasindhu Sarkar, Tanmay Shah, Gengyan Li, Daoye Wang, Leonhard Helminger, Sergio Orts-Escolano, Dmitry Lagun, Otmar Hilliges, Thabo Beeler, Abhimitra Meka
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 |
|---|---|---|---|
| initialise target latent by inversion rather than zeros, noise, mean or neighbour latents (ablation) | Preface model fitting · Studio dataset, 2 views, 1K | PSNR: 24.66 -> 25.69 | Improved |
| initialise from furthest training identity instead of inversion (ablation) | Preface model fitting · Studio dataset, 2 views, 1K | PSNR: 25.69 -> 23.91 | Hurt |
| add view-branch weight regulariser L_v (ablation) | Preface fine-tuning · Studio dataset, 2 views, 1K | PSNR: 23.91 -> 25.53 | Improved |
| add normal consistency loss L_normal (ablation) | Preface fine-tuning · Studio dataset, 2 views, 1K | PSNR: 25.53 -> 25.69 | Improved |
| increase number of fitting views from 2 to 7 (ablation) | Preface · Studio dataset, 1K | PSNR: 25.69 -> 29.24 | Improved |
| train prior on more identities (15 to 750) (ablation) | Preface prior · Studio dataset, 2 views, 1K | PSNR: 24.25 -> 25.43 | Improved |
| also optimise latent code during fine-tuning rather than model weights only (ablation) | Preface fine-tuning · Studio dataset, 2 views, 1K, subjects A/B/C | PSNR: 26.07 -> 26.55 (subject A) | Improved |
| use Preface prior instead of KeypointNeRF at 2 views | few-shot face NVS · Studio dataset, 2 views, 1K | LPIPS: 0.2713 -> 0.1905 | Improved |
| Preface (not trained on FaceScape) compared with DINER on FaceScape subject 344 | Preface · FaceScape, 4 views | PSNR: 25.51 vs 26.45 (DINER) | Mixed |
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.
- FreeNeRF — FreeNeRF: Improving Few-Shot Neural Rendering With Free Frequency Regularization (CVPR 2023) · 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
- PortraitSR: Artist-Inspired Prior Learning for Progressive Face Super-Resolution (AAAI 2026 · similarity 0.89)
- Face Super-Resolution Guided by 3D Facial Priors (ECCV 2020 · similarity 0.89)
- High Resolution Zero-Shot Domain Adaptation of Synthetically Rendered Face Images (ECCV 2020 · similarity 0.88)
- Learning Formation of Physically-Based Face Attributes (CVPR 2020 · similarity 0.88)
- CapHuman: Capture Your Moments in Parallel Universes (CVPR 2024 · similarity 0.88)
- Synthetic Prior for Few-Shot Drivable Head Avatar Inversion (CVPR 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.
