Research map / 3D vision / 3D generation
Monocular Avatar Reconstruction via Cascaded Diffusion Priors and UV-Space Differentiable Shading
Hong Li, Minqi Meng, Yanjun Liang, Chongjie Ye, Houyuan Chen, Weiqing Xiao, Xianda Guo, Guojun Lei, Xuhui Liu, Chaojie Yang, Yanlun Peng, HAO ZHAO and 1 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 |
|---|---|---|---|
| add multi-scale fusion with frozen DINOv3 to the ConvNeXt encoder (ablation) | geometry encoder · REALY | NMSE (mm): 1.806 -> 1.490 | Improved |
| add enhanced 88-point landmark loss (ablation) | geometry encoder · REALY mouth region | NMSE (mm): 1.670 -> 1.376 | Improved |
| add enhanced landmark loss (ablation) | geometry encoder · REALY overall | NMSE (mm): 1.518 -> 1.490 overall; mouth 1.670 -> 1.376; nose 1.457 -> 1.619; cheeks 1.140 -> 1.181 | Mixed |
| replace ConvNeXt baseline with full geometry model (ablation) | geometry encoder · REALY | NMSE (mm): 2.514 -> 1.490 | Improved |
| use linear 3DMM with DINOv3 priors vs nonlinear MoSAR | geometry · REALY | NMSE (mm): 1.500 -> 1.490 | Improved |
| compare with HiFace and 3DDFA-V3 | geometry · REALY | NMSE (mm): 1.275 / 1.435 vs 1.490 | Hurt |
| use Inpainting LoRA vs UV-IDM/HRN | texture inpainting · UV texture inpainting | PSNR: 19.39 -> 22.44 (vs UV-IDM) | Improved |
| use Light-Homogenization LoRA vs FFHQ-UV | light homogenization · light homogenization | CSIM: 0.1340 -> 0.4667 | Improved |
| add light homogenization stage before material estimation (ablation) | material estimation · avatar materials | material quality | Improved |
| use joint cross-intrinsic attention with differentiable BRDF loss instead of separate LoRAs (ablation) | material estimation · normal maps | qualitative | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- FFHQ-UV — FFHQ-UV: Normalized Facial UV-Texture Dataset for 3D Face Reconstruction (CVPR 2023) · better on 2 of 2 shared results
- MoSAR — MoSAR: Monocular Semi-Supervised Model for Avatar Reconstruction using Differentiable Shading (CVPR 2024) · better on 1 of 1 shared result
- HiFace — HiFace: High-Fidelity 3D Face Reconstruction by Learning Static and Dynamic Details (ICCV 2023) · better on 0 of 1 shared result
- Deep3D — Deep3D: Fully Automatic 2D-to-3D Video Conversion with Deep Convolutional Neural Networks (ECCV 2016) · 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
- GAF: Gaussian Avatar Reconstruction from Monocular Videos via Multi-view Diffusion (CVPR 2025 · similarity 0.90)
- Vid2Avatar-Pro: Authentic Avatar from Videos in the Wild via Universal Prior (CVPR 2025 · similarity 0.89)
- Human-3Diffusion: Realistic Avatar Creation via Explicit 3D Consistent Diffusion Models (NeurIPS 2024 · similarity 0.89)
- IntrinsicAvatar: Physically Based Inverse Rendering of Dynamic Humans from Monocular Videos via Explicit Ray Tracing (CVPR 2024 · similarity 0.89)
- Synthetic Prior for Few-Shot Drivable Head Avatar Inversion (CVPR 2025 · similarity 0.88)
- MoSAR: Monocular Semi-Supervised Model for Avatar Reconstruction using Differentiable Shading (CVPR 2024 · similarity 0.88)
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.
