Research map / 3D vision / Face and shape
FFHQ-UV: Normalized Facial UV-Texture Dataset for 3D Face Reconstruction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the FFHQ-UV-trained nonlinear texture decoder and reconstruction pipeline instead of the strongest reported baseline | 3D face reconstruction on REALY · REALY | all-region shape error: 1.868 -> 1.399 | Helped |
| Add multi-view texture extraction rather than using only a single frontal view (ablation) | FFHQ-UV reconstruction system · REALY | all-region shape error: 1.473 -> 1.399 | Helped |
| Replace the linear PCA texture basis with the learned texture decoder (ablation) | 3D face reconstruction · REALY | all-region shape error: 1.520 -> 1.399 | Helped |
| Pretrain on FFHQ-UV and finetune the decoder with Facescape rather than use the FFHQ-UV decoder alone (ablation) | 3D face reconstruction · REALY | all-region shape error: 1.399 -> 1.450 | Hurt |
| Increase texture-decoder training data from 5,000 to 54,165 maps (ablation) | StyleGAN2 texture decoder · CelebA-UV-100 | LPIPS error: 0.2029 -> 0.1487 | Helped |
| Add multi-view texture extraction rather than using only a single frontal view (ablation) | FFHQ-UV creation pipeline · FFHQ-UV | identity similarity: 0.7195 -> 0.7818 | Helped |
| Use UV-texture correction and completion instead of naive template blending (ablation) | FFHQ-UV creation pipeline · FFHQ-UV | identity similarity: 0.7712 -> 0.7818 | Helped |
| Add StyleGAN-based editing to normalize in-the-wild faces before texture extraction (ablation) | FFHQ-UV creation pipeline · FFHQ-UV | BS Error: 11.385 -> 7.293 | Helped |
| Use texture correction and completion rather than naive blending to the template (ablation) | FFHQ-UV creation pipeline · UV texture maps | visual artifact quality | Helped |
| Use multi-view rather than only frontal-view texture extraction (ablation) | FFHQ-UV creation pipeline · UV texture maps | actual facial texture coverage | Helped |
| Normalize in-the-wild images with StyleGAN-based editing before extracting textures (ablation) | FFHQ-UV creation pipeline · UV texture maps | illumination and occlusion quality | Helped |
Closest papers by meaning
- FreeUV: Ground-Truth-Free Realistic Facial UV Texture Recovery via Cross-Assembly Inference Strategy (CVPR 2025 · similarity 0.89)
- Towards High-Fidelity 3D Face Reconstruction From In-the-Wild Images Using Graph Convolutional Networks (CVPR 2020 · similarity 0.88)
- UV-IDM: Identity-Conditioned Latent Diffusion Model for Face UV-Texture Generation (CVPR 2024 · similarity 0.87)
- SFDM: Robust Decomposition of Geometry and Reflectance for Realistic Face Rendering from Sparse-view Images (CVPR 2025 · similarity 0.86)
- Blind Face Restoration under Extreme Conditions: Leveraging 3D-2D Prior Fusion for Superior Structural and Texture Recovery (AAAI 2024 · similarity 0.84)
- Text-Guided 3D Face Synthesis - From Generation to Editing (CVPR 2024 · similarity 0.84)
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.